SELF-RECOMMENDING!

Autism And Intelligence: Much More Than You Wanted To Know

[Thanks to Marco DG for proofreading and offering suggestions]

I.

Several studies have shown a genetic link between autism and intelligence; genes that contribute to autism risk also contribute to high IQ. But studies show autistic people generally have lower intelligence than neurotypical controls, often much lower. What is going on?

First, the studies. This study from UK Biobank finds a genetic correlation between genetic risk for autism and educational attainment (r = 0.34), and between autism and verbal-numerical reasoning (r = 0.19). This study of three large birth cohorts finds a correlation between genetic risk for autism and cognitive ability (beta = 0.07). This study of 45,000 Danes finds that genetic risk for autism correlates at about 0.2 with both IQ and educational attainment. These are just three randomly-selected studies; there are too many to be worth listing.

The relatives of autistic people will usually have many of the genes for autism, but not be autistic themselves. If genes for autism (without autism itself) increase intelligence, we should expect these people to be unusually smart. This is what we find; see Table 4 here. Of 11 types of psychiatric condition, only autism was associated with increased intelligence among relatives. This intelligence is shifted towards technical subjects. About 13% of autistic children (in this sample from whatever social stratum they took their sample from) have fathers who are engineers, compared to only 5% of a group of (presumably well-matched?) control children (though see the discussion here) for some debate over how seriously to take this; I am less sure this is accurate than most of the other statistics mentioned here.

Further (indirect) confirmation of the autism-IQ link comes from evolutionary investigations. If autism makes people less likely to reproduce, why would autism risk genes stick around in the human population? Polimanti and Gelemter (2017) find that autism risk genes aren’t just sticking around. They are being positively selected, ie increasing with every generation, presumably because people with the genes are having more children than people without them. This means autism risk genes must be doing something good. Like everyone else, they find autism risk genes are positively correlated with years of schooling completed, college completion, and IQ. They propose that the reason evolution favors autism genes is that they generally increase intelligence.

But as mentioned before, autistic people themselves on average have lower intelligence. One study found that 69% of autistic people had an IQ below 85 (the average IQ of a high school dropout). Only 3% of autistic people were found to have IQs above 115, even though 15% of the population should be at this level.

These numbers should be taken with very many grains of salt. First, IQ tests don’t do a great job of measuring autistic people. Their intelligence tends to be more imbalanced than neurotypicals’, so IQ tests (which rely on an assumption that most forms of intelligence are correlated) are less applicable. Second, even if the test itself is good, autistic people may be bad at test-taking for other reasons – for example, they don’t understand the directions, or they’re anxious about the social interaction required to answer an examiner’s quetsions. Third, and most important, there is a strong selection bias in the samples of autistic people. Many definitions of autism center around forms of poor functioning which are correlated with low intelligence. Even if the definition is good, people who function poorly are more likely to seek out (or be coerced into) psychiatric treatment, and so are more likely to be identified. In some sense, all “autism has such-and-such characteristics” studies are studying the way people like to define autism, and tell us nothing about any underlying disease process. I talk more about this in parts 2 and 3 here.

But even adjusting for these factors, the autism – low intelligence correlation seems too strong to dismiss. For one thing, the same studies that found that relatives of autistic patients had higher IQs find that the autistic patients themselves have much lower ones. The existence of a well-defined subset of low IQ people whose relatives have higher-than-predicted IQs is a surprising finding that cuts through the measurement difficulties and suggests that this is a real phenomenon.

So what is going on here?

II.

At least part of the story is that there are at least three different causes of autism.

1. The “familial” genes mentioned above: common genes that increase IQ and that evolution positively selects for.

2. Rare “de novo mutations”, ie the autistic child gets a new mutation that their non-autistic parent doesn’t have. These mutations are often very bad, and are quickly selected out of the gene pool (because the people who have them don’t reproduce). But “quickly selected out of the gene pool” doesn’t help the individual person who got one of them, who tends to end up severely disabled. In a few cases, the parent gets the de novo mutation, but for whatever reason doesn’t develop autism, and then passes it onto their child, who does develop autism.

3. Non-genetic factors. The best-studied are probably obstetric complications, eg a baby gets stuck in the birth canal and can’t breathe for a long time. Pollution, infection, and trauma might also be in this basket.

These three buckets and a few other less important factors combine to determine autism risk for any individual. Combining information from a wide variety of studies, Gaugler et al estimate that about 52% of autism risk is attributable to ordinary “familial” genes, 3% to rare “de novo” mutations, 4% to complicated non-additive genetic interaction effects, and 41% “unaccounted”, which may be non-genetic factors or genetic factors we don’t understand and can’t measure. This study finds lower heritability than the usual estimates (which are around 80% to 90%; the authors are embarrassed by this, and in a later study suggest they might just have been bad at determining who in their sample did or didn’t have autism. While their exact numbers are doubtful, I think the overall finding that common familial genes are much more important than rare de novo mutations survives and is important.

Most cases of autism involve all three of these factors; that is, your overall autisticness is a combination of your familial genes, mutations, and environmental risk factors.

One way of resolving the autism-intelligence paradox is to say that familial genes for autism increase IQ, but de novo mutations and environmental insults decrease IQ. This is common-sensically true and matches previous research into all of these factors. So the only question is whether the size of the effect is enough to fully explain the data – or whether, even after adjusting out the degree to which autism is caused by mutations and environment, it still decreases IQ.

Ronemus et al (2014) evaluate this:

They find that even autistic people without de novo mutations have lower-than-average IQ. But they can only screen for de novo mutations they know about, and it could be that they just missed some.

Here’s another set of relevant graphs:

This one comes from Gardner et al (2019), which measures the cognitive ability of the fathers of autistic people and disaggregates those with and without intellectual disability. In Graph A, we see that if a child has autism (but not intellectual disability), their likelihood of having a father with any particular IQ (orange line) is almost the same as the likelihood of a neurotypical child having a father of that IQ (dotted line). Disguised in that “almost” is a very slight tendency for fathers to be unusually intelligent, plus a (statistically insignificant) tendency for them to be unusually unintelligent. For reasons that don’t entirely make sense to me, if instead we look at the likelihood of the father to be a certain intelligence (bottom graph, where dark line surrounded by gray confidence cloud is autistic people’s fathers, and dotted line is neurotypical people’s fathers) it becomes more obvious that more intelligent people are actually a little more likely to have autistic children (though less intelligent people are also more likely.

(remember that “no intellectual disability” just means “IQ over 70”, and so many of these not-intellectually-disabled people may still have low intelligence – I wish the paper had quantified this)

Graph B is the same thing, but with people have have autism with intellectual disability. Now there is a very strong effect towards their fathers being less intelligent than usual.

This confuses me a little. But for me the key point is that high-intelligence fathers show a trend (albeit not significant in this study) to be more likely than average to have children with autism and intellectual disability.

These questions interest me because I know a lot of people who are bright nerdy programmers married to other bright nerdy programmers, and sometimes they ask me if their children are at higher risk for autism. While their children are clearly at higher risk for autistic traits, I think they want to know whether they have higher risk for the most severe forms of the syndrome, including intellectual disability and poor functioning. If we take the Ronemus and Gardner studies seriously, the answer seems to be yes. The Gardner study seems to suggest it’s a very weakly elevated risk, maybe only 1.1x or 1.2x relative risk. But the Gardner study also ceilings off at 90th percentile intelligence, so at this point I’m not sure what to tell these people.

III.

If Ronemus isn’t missing some obscure de novo mutations, then people who get autism solely by accumulation of common (usually IQ-promoting) variants still end up less intelligent than average. This should be surprising; why would too many intelligence-promoting variants cause a syndrome marked by low intelligence? And how come it’s so inconsistent, and many people have naturally high intelligence but aren’t autistic at all?

One possibility would be something like a tower-vs-foundation model. The tower of intelligence needs to be built upon some kind of mysterious foundation. The taller the tower, the stronger the foundation has to be. If the foundation isn’t strong enough for the tower, the system fails, you develop autism, and you get a collection of symptoms possibly including low intelligence. This would explain low-functioning autism from de novo mutations or obstetric trauma (the foundation is so weak that it fails no matter how short the tower is). It would explain the association of genes for intelligence with autism (holding foundation strength constant, the taller the tower, the more likely a failure). And it would also explain why there are many extremely intelligent people who don’t have autism at all (you can build arbitrarily tall towers if your foundation is strong enough).

I’ve only found one paper that takes this model completely seriously and begins speculating on the nature of the foundation. This is Crespi 2016, Autism As A Disorder Of High Intelligence. It draws on the VPR model of intelligence, where g (“general intelligence”) is divided into three subtraits, v (“verbal intelligence”), p (“perceptual intelligence”), and r (“mental rotation ability”) – despite the very specific names each of these represents ability at broad categories of cognitive tasks. Crespi suggests that autism is marked by an imbalance between P (as the tower) and V + R (as the foundation). In other words, if your perceptual intelligence is much higher than your other types of intelligence, you will end up autistic.

It doesn’t really present much evidence for this other than that autistic people seem to have high perceptual intelligence. Also, it doesn’t really look like autistic people are worse at mental rotation. Also, the Gardner paper has analyzed autistic patients’ fathers by subtype of intelligence, and there is a nonsignificant but pretty suggestive tendency for them to have higher-than-normal verbal intelligence; certainly no signs of high verbal intelligence preventing autism. I can’t tell if this is evidence against Crespi or whether since all intellectual abilities are correlated this is just the shadow of their high perceptual intelligence, and if we directly looked at perceptual-to-verbal ratio we would see it was lower than expected. Also also, Crespi is one of those scientists who constantly has much more interesting theories than anyone else (eg), and this makes me suspicious.

Overall I would be surprised if this were the real explanation for the autism-and-intelligence paradox, but it gets an A for effort.

Conclusions

1. The genes that increase risk of autism are disproportionately also genes that increase intelligence, and vice versa (~100% confidence)

2. People diagnosed with autism are less intelligent than average (~100% confidence, leaving aside definitional complications)

3. Some of this effect is because autism is caused both by normal genes and by de novo mutations and environmental insults, and the de novo mutations and environmental insults definitely decrease intelligence. Every autism case is caused by some combination of these three factors, and the more it is caused by normal genes, the more intelligence is likely to be preserved (~100% confidence)

4. This is not the whole story, and even cases of autism that are caused entirely or mostly by normal genetics are associated with unusually low IQ (80% confidence)

5. This can best be understood through a tower-versus-foundation model where higher intelligence that outstrips the ability of some mysterious foundation to support it will result in autism (25% confidence)

6. The specific way the model plays out may be through perceptual intelligence out of balance with verbal and rotational intelligence causing autism (3% confidence)

This entry was posted in Uncategorized and tagged , , , . Bookmark the permalink.

Leave a Reply

213 Responses to Autism And Intelligence: Much More Than You Wanted To Know

  1. Benito says:

    (The post is awesome, here’s a few errata, feel free to delete this after reading.) The line here has an extra close-paragraph symbol

    > (though see the discussion here) for some debate over how seriously to take this; I am less sure this is accurate than most of the other statistics mentioned here)

    which I assume is the one missing from the later section here

    > (which are around 80% to 90%; the authors are embarrassed by this, and in a later study suggest they might just have been bad at determining who in their sample did or didn’t have autism. While their exact numbers are doubtful, I think the overall finding that common familial genes are much more important than rare de novo mutations survives and is important.

    Also, “IQ over 70” should be ““IQ under 70”, and “people have have autism” should be “people who have autism.

    • melolontha says:

      More errata, I think:

      This one comes from Gardner et al (2019), which measures the cognitive ability of the fathers of autistic people and disaggregates those with and without intellectual disability. In Graph A, we see that if a child has autism (but not intellectual disability), their likelihood of having a father with any particular IQ (orange line) is almost the same as the likelihood of a neurotypical child having a father of that IQ (dotted line). Disguised in that “almost” is a very slight tendency for fathers to be unusually intelligent, plus a (statistically insignificant) tendency for them to be unusually unintelligent. For reasons that don’t entirely make sense to me, if instead we look at the likelihood of the father to be a certain intelligence (bottom graph, where dark line surrounded by gray confidence cloud is autistic people’s fathers, and dotted line is neurotypical people’s fathers) it becomes more obvious that more intelligent people are actually a little more likely to have autistic children (though less intelligent people are also more likely.

      I don’t think this paragraph fits the note in the image. First, doesn’t the note say that the bottom graph (for each letter) is the likelihood that a father of x IQ will have a child with the specified outcome? And second, doesn’t it say that in the bottom graph, the dotted and solid lines represent unadjusted and adjusted estimates, respectively?

  2. Soy Lecithin says:

    I’ve heard stories of autistic children sometimes regressing in development around 2 years old, e.g. children who previously had normal or even remarkable verbal abilities for their age suddenly “losing words”. Maybe this is a “tower collapse” event.

    • spandrel says:

      This pattern is not uncommon (my son followed this trajectory), and is partly responsible for the even less uncommon belief that vaccines cause autism, because the regression coincides with vaccination.

    • chridd says:

      Other possible explanation(s): Some kind of anxiety about doing things wrong, combined with worrying about whether one is “supposed” to be able to do a thing yet. As an example of the latter… there was a time when I was in math class when we’d started learning about simple equations with variables (4n + 2 = 3n + 5 or whatever) but we hadn’t yet learned about adding things to both sides (I think we were just supposed to guess and check). I’d figured out myself how to do that, but tried to hide that I knew how because I thought I wasn’t supposed to know yet. (Then also add to that the fact that people making a big deal out of “Wow you can do this thing!” can be super unpleasant…)

      • Ben Wōden says:

        In primary school, for a while we used to have daily spelling tests, where we had to spell ten words. If we got 10/10, we’d get a gold star, and, if 9/10, a silver star. I always used to get one word wrong on purpose because I preferred the silver stars. I still sometimes feel bad about whether that ended up slightly ruining some sort of investigation into which methods of teaching spelling were better or something like that. It would not surprise me if a lot of non-neurotypical children were doing similar things, and that it poses genuine problems for studies like those discussed above.

  3. I’ve only found one paper that takes this model completely seriously and begins speculating on the nature of the foundation. This is Crespi 2016, Autism As A Disorder Of High Intelligence. It draws on the VPR model of intelligence, where g (“general intelligence”) is divided into three subtraits, v (“verbal intelligence”), p (“perceptual intelligence”), and r (“mental rotation ability”) – despite the very specific names each of these represents ability at broad categories of cognitive tasks. Crespi suggests that autism is marked by an imbalance between P (as the tower) and V + R (as the foundation). In other words, if your perceptual intelligence is much higher than your other types of intelligence, you will end up autistic.

    It doesn’t really present much evidence for this other than that autistic people seem to have high perceptual intelligence. Also, it doesn’t really look like autistic people are worse at mental rotation. Also, the Gardner paper has analyzed autistic patients’ fathers by subtype of intelligence, and there is a nonsignificant but pretty suggestive tendency for them to have higher-than-normal verbal intelligence; certainly no signs of high verbal intelligence preventing autism. I can’t tell if this is evidence against Crespi or whether since all intellectual abilities are correlated this is just the shadow of their high perceptual intelligence, and if we directly looked at perceptual-to-verbal ratio we would see it was lower than expected. Also also, Crespi is one of those scientists who constantly has much more interesting theories than anyone else (eg), and this makes me suspicious.

    Interesting. For what it’s worth, I was diagnosed with Aspergers (which has now been folded into autism) at age 12 and as part of the diagnosis they did an IQ test with a “processing IQ” component and a “verbal IQ” component, and although the documents I have don’t list the actual scores, the result is referred to in the documents as being very high in the verbal component and very low in the processing component.

  4. Thegnskald says:

    Has anyone here actually met a highly intelligent person who is genuinely normal? Not just running an emulation of “normal” to deal with other people?

    • Michael Handy says:

      Yes. Generally high end career civil servants and similar.

      Terrifyingly competent and motivated, shockingly good with people, and with few visible mental issues. It’s possible they just have a very good emulation that never slips even in casual situations, but at some point that basically merges with normal.

      • aristides says:

        This is interesting to me, since I am on a career path to become a high end career civil servant, GS-12 at 25 years old for those that know what that means, and I am anything but normal. I’m somewhere on the autism spectrum and have high levels of anxiety. I am able to emulate being normal very well at work, but horrible in casual settings. This has held me back at networking events, but overall, there is some tolerance for being abnormal out of the office. I would say half of the up and coming civil leaders have hints of being abnormal. What will be interesting to see if we get weeded out before getting close to the top, and if the breakdown will stay the same. Fortunately, at home, I can act how I want, which recharges me significantly.

        • JohnofCharleston says:

          Strongly agree. GS-14 speaking, and I’m not exactly normal either. The best civil servants get an extra allowance of weirdness points to allocate as necessary.

      • Thegnskald says:

        From a certain perspective, it merges with normal.

        From another…

        If I suggest autism is effectively what we call people whose… mental resolution (in the sense of pixels per inch) is too high, would that make sense to you?

        • LadyJane says:

          This seems like circular logic to me. By the same rationale, you could say that no one is normal, and everyone is simply running a mental emulation.

          You’re starting with the premise “intelligence is incompatible with normality” and then reflexively dismissing all evidence to the contrary with “they’re not really normal, they’re just pretending.” Which is 1.) completely unfalsifiable, and 2.) could equally be applied to people who aren’t especially intelligent.

          • Thegnskald says:

            This is true in a sense, but also alters what we mean by “normal”.

            Normal is being concerned enough with appearing normal to try being normal, yes.

            I notice a lot of people confuse normality with charisma. Having charismatic superpowers isn’t “normal”, and I suspect requires being sufficiently unconcerned with opinion – sufficiently thing-oriented – so as to cease to be normal.

          • LadyJane says:

            How exactly are you defining “normal”? A lot of people are using social aptitude as a proxy because that’s at least something measureable (at least in concept, if not in practice). Normality is just too vague to be quantifiable.

            Your argument also seems tautological to an extent; if you’re defining “normal” as “someone who thinks like a normal person,” that’s basically the equivalent of describing an object as being shaped like itself. It also becomes meaningless to say that “no highly intelligent people are normal,” since by definition, highly intelligent people will think in different ways than people with average intelligence.

            You could also define “normal” as simply referring to a lack of explicit mental illnesses or emotional problems. But then the answer to your question becomes so obvious that it’s hardly worth asking in the first place; there are countless people who are highly intelligent without suffering from any kind of psychological condition or personality disorder or mental/emotional instability.

        • Meister says:

          autism is effectively what we call people whose… mental resolution (in the sense of pixels per inch) is too high

          I had a professor once whose research thesis was basically this applied to language. The idea is that autistic children have more and finer-grained category boundaries, causing them to over-differentiate similar phonemes that most speakers use interchangeably. She hoped to use this theory to explore why some autistics show delayed language acquisition.

      • Clutzy says:

        Umm, what? I used to live in the DC area and still work with the civil service in my current job.

        Obviously it cannot be 100%, but I have not yet encountered one that isn’t a narcissist or what the internet calls an “autist”. James Comey’s twitter and book are exactly what I’d expect from a career civil service person. Naive, narcissistic, completely detached from human interaction. These are the people that make the Turing test invalid, you’d think they were the chatbot.

        • Simon_Jester says:

          Would you mind clarifying what you mean by “narcissist?” I mean, if you go with The Last Psychiatrist’s standards we are basically all narcissists and to some extent the claim “Comey is a narcissist” loses meaning.

          • LadyJane says:

            The bit about Comey honestly just seems like partisan sniping to me. The odds of him actually having Narcissistic Personality Disorder or any kind of Autistic Spectrum Disorder are close to 0. It’s highly unlikely for someone with either of those conditions to become the Director of the FBI.

          • Clutzy says:

            I am not saying they have the full disorder (and no its not a partisan snipe, except to the extent that being very skeptical of the US bureaucratic class is partisan). I am simply saying that when you encounter lots of GS11+ people they are exactly like Comey. That’s probably how he got to that level. He is the most Civil Servicy person around. They get offended more if you question them than normal lawyers do, they have the 4chan-y social awkwardness that those communities jokingly call “autism”, etc.

            Its a highly abnormal subset of people. Regarding Narcissim, the first website I pulled up had these as the first 3 symptoms:

            * Have an exaggerated sense of self-importance
            * Have a sense of entitlement and require constant, excessive admiration
            * Expect to be recognized as superior even without achievements that warrant it

            These 3 have a near 100% incidence rate among civil service employees that I’ve had to interact with.

    • Chalid says:

      People at the higher end of the corporate ladder that I’ve met were very intelligent and also seemed generally normal, though of course they had much higher-than-normal ambition, stamina, and interest in their business.

    • Saint Fiasco says:

      Beware Simpson’s Paradox. Plenty of people are both smarter and more charismatic than us. We just rarely see them because they don’t hang out with us losers.

      Within our filter bubble, people have similar levels of status, so they can be smart but not charismatic or charismatic but not smart. We don’t let people who are neither smart nor charismatic hang out with us, and people who are both smart and charismatic don’t want to hang out with us.

      • Thegnskald says:

        You underestimate the potential of a good emulation, I think.

        • Simon_Jester says:

          I’m pretty sure that at some point the emulation efficiency hits diminishing returns- that it’s not “emulations all the way up,” in other words.

          There’s no reason to assume that every intelligence-enhancing trait necessarily increases “weirdness.” And so, we’d expect that not every person blessed with a strong combination of intelligence-enhancing traits is going to have an excess of weirdness that they have to brute-force emulate their way around.

          Some, yes; all, no.

          • Shion Arita says:

            Well, I think it’s certainly a reasonable possibility, which in my experience appears to be true, that beyond a certain point, intelligence either directly causes ‘weirdness’ or is reliant upon it as a foundation. Either way, I don’t think you can have one without the other above a certain level.

            I guess it depends on how you define ‘weirdness’; I define it as “having interests, beliefs, or though processes that are significantly different than most people” I don’t think I’ve encountered any extremely intelligent people for whom none of the above are the case.

          • Simon_Jester says:

            If your definition of ‘weirdness’ includes beliefs and interests, then almost by definition anyone whose underlying cognition is unusually effective or ineffective will be ‘weird.’ Thus, “I can’t think of anyone who’s very intelligent and not weird” ceases to be meaningful.

            For purposes of a discussion of this, I think it’s best if we restrict ourselves to unusual behaviors. Insofar as autism has a coherent definition, it is a category that embraces a variety of behavioral and neural processing anomalies.

            Thus, the question under discussion becomes “does every highly intelligent person have some of these autism-linked genes, or are there other paths to high intelligence that do not involve these genes?”

      • Bugmaster says:

        Agreed. There are plenty of incredibly smart and simultaneously charismatic people out there. There are maybe as many as 1000 of them in the Western world, and none of them would want to ever hang out with us — because they are smart enough to know that their time would be better spent managing their billion-dollar corporations and/or governments.

        • LadyJane says:

          That figure seems very low to me, though it depends what exactly you define as “incredibly smart” and “incredibly charismatic.” If we take “incredibly” to be synonymous with top percentile, there would be 75 million people in the world who are incredibly intelligent. Assuming that intelligence is neither positively nor negatively correlated with social skills, that means 1% of those 75 million people would also be in the top percentile for social aptitude, for a total of 750,000 people. The Western world (taken to mean Europe and the Anglosphere) comprises roughly 1/6th of the global population, which means there should be about 125,000 Westerners who are both incredibly smart and incredibly charismatic. Of course, since intelligence and social aptitude seem to have a slight positive correlation, the real number might be closer to 150,000 or 200,000.

          • Rachael says:

            I think top 1% is too low a bar for “incredibly”. Someone at the 99.0 percentile for intelligence would probably seem quite mediocre by SSC standards.

          • LadyJane says:

            Someone at the 99.0 percentile for intelligence would probably seem quite mediocre by SSC standards.

            I strongly doubt that. And yes, I know the survey showed that the average SSC reader had an IQ of 138, but I’m extremely skeptical of that finding too. I suspect a lot of people who answered the survey were giving scores from IQ tests they took as children, or from internet tests of dubious quality. And at least a few of them were probably just lying outright.

            But if you want to redefine “incredible” to mean the top 0.1 percent of the population, then yes, Bugmaster’s guess would be roughly accurate: There’d be approximately 1,250 people who were both incredibly smart and incredibly charismatic in the Western world, or maybe 1,500-2,000 if you assume a minor positive correlation between being intelligence and being socially adept.

          • quanta413 says:

            I think somewhere in the 90s percentile of IQ is plausibly normal by SSC standards.

            I don’t think 99% would stand out here, but I don’t think it’d quite be mediocre either.

            The differences between 99% and 99.5% get swamped by other differences between people within that group in traits like energy or perseverance or whatever. What will really make someone come off as intelligent within that group is going to be something like 99% IQ plus 99% along that other stuff. Energy/perseverance/drive and other harder to measure factors.

        • Thegnskald says:

          Why do people keep thinking charisma is normality, out or curiosity? Is it because that is a trait many people here associate with passing for normal?

    • AlexanderTheGrand says:

      I’m really curious, how many people identify with this statement? I feel like unless your definition of “normal” includes “non-highly intelligent” this is just so far from true.

      I don’t know their actual IQs, but I’m very confident a lot of people in my bubble are nice, funny, enjoyable, sports-watching people with 99% intelligence.

      Maybe I just don’t understand what you mean by normal. Could you elaborate?

      • Thegnskald says:

        Hrm. Perhaps it will help if I ask a question.

        How much effort do people, particularly young people, put into being “normal” – that is, into fitting in?

        • AlexanderTheGrand says:

          Maybe one bias from my statement is I’m young, and many of my smart friends are also young. I imagine the effort you put into fitting in goes down as you get older. Good clarifying question.

          I’m going to interpret “fitting in” as “being liked.”

          A lot of the super smart people I know also seem super interesting — either being very funny, uniquely talented, or engaging. The talented ones don’t seem to put extra effort into being impressive for other’s sake. The other two options I guess take effort, but I’d say it’s effort that they enjoy putting in, because they enjoy the rewards (friendship, novel socializing, etc.).

          So in that sense, maybe it’s just a phrasing problem. Do a lot of super smart people play the game? I’d say so. But many do it for positive outcomes versus negative pressure, which is how I interpreted your question.

          • Thegnskald says:

            Hrm. I think you’d need to ask them about their internal experiences, but I’d guess a lot of the people you find unusually functional aren’t putting a lot of effort into fitting in.

            Struggling to describe this. Normal includes insecurities, anxieties, caring what other people think. Take these away and you arrive at something that seems like an effortless confidence and freedom. I’d guess the funny people can be funny even when their jokes bomb, because they can make their failed jokes into a joke at their own expense?

      • Dammit, I was nodding along until you threw in sports-watching.

        (:D Sorry, couldn’t resist the quip; please carry on.)

    • I’m not sure what you mean by normal. I’ve certainly known highly intelligent people who did not seem to be crazy or disfunctional in any obvious way. Highly intelligent people are likely to have different interests, read and talk about different things, than average people–does that make them not normal?

      • Bugmaster says:

        FWIW, I know a few bona-fide genius-level people; on the order of “Ph.D. in particle physics at age 15, millionaire at age 30” type of stuff. By and large, they find conversations with non-genius people extremely boring, because they can predict exactly what the other person will say long before he says it. For this reason, they don’t tend to hang out with normal people much.

        • jefftk says:

          You might want to adjust for inflation. A PhD in particle physics at 15 is incredibly impressive, but a programmer starting out of college at an FAANG who saves a bit will be a millionaire by 30.

        • Cliff says:

          Are you sure any of that is a real thing?

          • Edward Scizorhands says:

            If you are an expert in X, and the popular understanding of X is X’, you will end up having the exact same conversation about X every time. It would be an excellent skill for a teacher.

            (I have “predicted” conversations, but only because the other person forgot we already had it. I then play out both parts of the conversation from memory as their mouth gapes open.)

        • anonymousskimmer says:

          Honestly, your genius doesn’t sound very good at digging into people. He (I presume) seems to be a bad conversationalist, likely through lack of practice. It’s not difficult to guess what someone’s response would be to many chit-chat topics.

          This reminds me of something a past friend mentioned. This friend had been working for many years with Leroy Hood. During a conversation on something or other he told my friend that he knew what my friend was going to say, then ended up saying “that’s not what I’d thought you’d say” afterward. Hood got his M.D. at 25/26 and Ph.D. at 29/30, so might not be of the caliber of your acquaintance, though.

      • Thegnskald says:

        Hard to specify. “I know it when I see it” and all that.

        I guess in a context-specific sense, normal is being defined as not-autistic?

    • There was definitely a sperg-normie dynamic on my last work team, all highly intelligent. The normies were not spergs running emulations, they’d have the same confused look whenever we sperged-out about something that we had when they talked about sportsball.

    • The Nybbler says:

      Sure, most of them. If you mean extraverted, no small number.

    • LadyJane says:

      Of course people like that exist. Intelligence and social aptitude don’t seem to be positively correlated, at least not very strongly, but they’re definitely not negatively correlated. If anything, there’s a slight positive correlation between the two. (That said, it becomes an incredibly strong correlation if you’re taking mentally disabled people into account, as people with IQs below 70 tend to have extremely poor social skills.)

      As for my own personal experience, I’ve found that people with high conceptual intelligence (as opposed to mathematical intelligence) tend to be more charismatic. I’ve known people who were very mathematically and/or technically inclined who had poor social skills. But the ability to notice patterns in the world, analyze phenomena and their causes/effects, understand highly complex and abstract concepts, and make connections between things all seem to be correlated with higher levels of social awareness and persuasive ability.

    • albatross11 says:

      I work with with a lot of very smart people, and have throughout my career. Most are weird in the sense of having uncommon interests (reading history or science books or studying languages for fun, say), but most seem pretty normal in terms of personality and affect and such. I’d say the people in my field are shifted about half a sigma toward introverted, toward socially awkward, and toward tending to become obsessive about their interests/hobbies, but that leaves a lot of people who’d seem pretty normal in most ways. My office mate is a big sports fan who cares a lot about his yard and house and is a fairly partisan Democrat…and who’s also a mathematician who plays folk music and speaks several languages. Most of my coworkers (except for the youngest ones) are married, and many have children. They’re often chatting about the latest sportsball game or what their kids are doing or their vacation plans–conversations that wouldn’t seem unusual anywhere in the US. Other times, the conversations would seem weird as hell in most places–discussions of physics or evolution or game theory or genetics or whatever, with a high background level of assumed knowledge.

    • Anthony says:

      Most successful businessmen. My current boss. My bosses at my last job (normal modulo their ethnic origin). A couple of property developers I’ve worked with.

      I’m not sure how “normal” she is, but an acquaintance through RenFaire is head of some Army Corps of Engineers office in the Bay Area, and she’s one of the very few people I consider noticeably more intelligent than me.

    • dyfed says:

      Most genuinely intelligent people are also fairly charismatic. Yes their interests and opinions are often eccentric, but they are self-aware of this and do not have problems navigating everyday interactions.

      People running ‘an emulation of normal’ are usually not staggeringly intelligent. Social relationships are not difficult and people who cannot navigate them are displaying a serious deficit. They may also have eccentric opinions or interests but they’re more likely to pursue them obsessively or compulsively without good reason, and they have trouble with self-awareness and awareness of others.

      Preoccupation with what’s ‘normal’ and whether you are ‘normal’ is a sign of mediocrity in my experience (though it’s not terminal). Brilliant, productive people don’t spend time either wishing to be normal, wishing to be different, or obsessing how they are different from ‘normal.’ They’re simply too busy getting things done for a triviality that reduces to dumb averages to matter to them.

  5. Le Maistre Chat says:

    If the IQ/autism risk genes are being selected for, we should be seeing more top-level geniuses each generation (I’m very skeptical of this) and more autistic children (this half checks out).

    • broblawsky says:

      That sounds like the Flynn effect, although the Flynn effect generally isn’t considered genetic.

    • Simon_Jester says:

      Except that from the sound of it, the IQ/autism risk genes are…

      [DANGER, OVERSIMPLIFICATION FOLLOWS]

      …a sort of arrangement where if you pick up to four of them you get enhanced intelligence, five or six and you get high-functioning intelligent autism, and seven or more and your brain crashes on boot-up and the autism stops you from functioning.

      If so, then the top-level geniuses may well have something else going on upstairs to enhance their intelligence, some other difference from the norms, that isn’t part of the complex of IQ/autism risk genes. Autism-risk genes may just be plumping up the middle tiers, those who are noticeably intelligent but not staggering one-in-a-million test cases.

      • albatross11 says:

        So one model is that we should see two things happening at once, as a result of the assortative mating by IQ that’s been going on for a couple generations now in the US:

        a. Increasingly smart people at the high end, as the smart people tend to meet in college/at work and pair off and have super-smart kids.

        b. Increasing numbers of autistic kids at the high end, as some of the smart kids get too many overlapping IQ-enhancing genes that mess up something important.

  6. martinw says:

    Any chance it’s just a matter of recessive genes? Like, if you have one copy of a particular high-IQ gene you get high IQ, but two copies of that gene is too much of a good thing and you get autism?

  7. broblawsky says:

    How much of this could be IQ tests failing to effectively measure g in autistic individuals?

    • ArawnGraalrd says:

      I would expect a challenge, in testing the intelligence of Autistics. It’s not like they’ll answer anybody. They would more easily get high IQ amongst Aspies, whom I don’t consider Autistic.

    • albatross11 says:

      There was a Conversations With Tyler podcast with Michelle Dawson (link here including transcript) where this was discussed. She pointed out that there are a couple widely-used kinds of IQ test. For one, it’s multiple different kinds of mental tasks whose scores are then combined to get an IQ score. For another (Ravens), it’s a bunch of iterations of the same basic task (a kind of visual puzzle with one piece missing), described here.

      Her claim was that autistics often did *way* better on Ravens than on the multiple subtest kind of IQ test. As I understand it (I’m an interested amateur, not an expert, so don’t trust this too far!), in the general population, you will see a very strong correlation between the IQ score someone gets on the two different kind of IQ test, but among autistics, the correlation is much weaker. (And maybe there’s some kind of subtype of autism –> size of correlation thing happening, too.)

  8. clipmaker says:

    There is a (high functioning) autistic musician named Henny Kupferstein who spends a lot of time teaching piano to autistic kids. She has absolute pitch, an attribute that is quite rare in the general population, but she says an amazingly high proportion of her autistic students have it to some degree, even when they didn’t know it. She has an article on her website here: https://hennyk.com/autism-and-perfect-pitch/ and a book: https://hennyk.com/book-perfect-pitch-in-the-key-of-autism/

    Some of the stuff she says about this seem a bit woo-woo, but interesting anyway. I don’t know what to make of this:

    Underconnectivity in autistic brains account for sensory integration differences and gifts. Sights and sounds are processed in the brain area that first receives it without sending it further to attach meaning. For such individuals, perfect pitch is almost always the indicator that highly refined processing is available in these areas of immediate reception of sound. This makes a very strong auditory learner. Auditory learning is therefore the only accessible means for learning prior to the bridging of other brain functions. This article explores piano lessons for non-verbal and autistic students as a means for creating new and necessary pathways for abstract logic inherent in higher education.

    • Ketil says:

      She has absolute pitch, […] an amazingly high proportion of her autistic students have it to some degree

      I see what you did there.

      • noyann says:

        Precision of it? Awareness of it? Moments of it?

      • jasmith79 says:

        There are different levels of the concept we call “perfect pitch”. Some people can correctly identify the pitch of a single note played on the piano. Go up a level and they can tell you the pitch of your hair dryer. Go up a level and you can mash a bunch of sequential notes on a piano with one left out of the middle and they can tell you the highest note, lowest note, and the one you left out. When I was a music major in college it was the research interest of one of my professors and he put the students with perfect pitch through a couple of tests like this in class.

        • Ben Wōden says:

          And then there’s Jacob Collier, who can detect, and accurately sing, the difference between an A at 440Hz and at 432Hz, and that allows him to do stuff like this (https://www.youtube.com/watch?v=mPZn4x3uOac) where he sings a 10-part harmonisation of “In the Bleak Midwinter” in which he’s gradually shifting the key-centre upwards throughout, and then for the last verse he performs a 4-chord pivot to a key halfway between G and G# on the scale he’s in at the time, by taking advantage of slight mis-definition in the equal-tempered approximations to those pivot chords that allows him to quickly shift to an entirely out-of-tune system (if you switched directly, it would just sound like trash) in just four chords without it sounding odd. It’s talked through in detail here (https://www.youtube.com/watch?v=Xd54l8gfi7M).

  9. zluria says:

    Couldn’t this all be the result of several essentailly different things being lumped together under the label “autism spectrum”? Perhaps the mechanisms for high IQ autism and low IQ autism are different.

    To test this, you could see if higher IQ autistic people tend to have more or fewer of these genes correlated with high intelligence, compared to low IQ autistic people. According to the “different mechanism” theory, they should have more, but by the “collapsing tower” theory, they should have fewer.

    • Well, much debate here gets into the distinction between people with “high functioning autism” and Aspergers, due to a muddling at that end of the scale. Aspergers was definitionally distinguished from autism due to a lack of the cognitive deficits associated with that condition, but in 2013, the DSM-5 folded it into autism spectrum disorders, so this could make it difficult to analyse the genes of high function autistics if that was the mistake I believe it was.

      It’s possible some degree of politics was involved in this decision:

      The World Health Organization also is eliminating Asperger syndrome from its International Classification of Diseases. The ICD-11, which was adopted this year and will be implemented globally by 2022, instead calls it “autism spectrum disorder without disorder of intellectual development and with mild or no impairment of functional language.” Proponents of the change hope to reduce stereotypes. For example, girls were much less likely to receive an Asperger’s diagnosis than boys, and girls were more likely to be diagnosed at an older age—a disparity that points to bias. Meanwhile, placing people “on the spectrum” equalizes access to resources, including insurance coverage.

    • arbitraryvalue says:

      several essentailly different things being lumped together under the label “autism spectrum”

      I suspect that this is the case. My experience watching a relative currently diagnosed with severe, non-verbal autism grow up is that his diagnosis is a catch-all for “he doesn’t seem capable of abstract thought but we have ruled out all the other, better-understood potential causes.” No one can offer more than conjecture about the cause of his condition, and no one knows how to treat it (unless you listen to the vaccine-chelation charlatans the way that his father does) so why is the medical establishment making confident claims regarding the connection between his condition and “high-functioning autism”?

      • Aapje says:

        why is the medical establishment making confident claims regarding the connection between his condition and “high-functioning autism”?

        The reason seems to be that there is great overlap in symptoms, although there is great diversity in symptoms.

  10. gkai says:

    Has paternal/maternal age been eliminated as cofounding factor from parents IQ/autism correlation? This seems a possible explanation, if high-IQ parents have children significantly later…

    I like the tower explanation, but not so much the different types of intelligences: maybe autism is a combination of rare IQ-boosting alleles inherited from parents, together with high general genetic load (which, among other detrimental effects, decrease IQ, negating the IQ-boosting alleles). the genetic load would be due to bad luck in parent mixing (recessive-recessive pairs) and parent age. Without the load, the child would have had a high IQ, but with weak fundation (genetic load) no. Still, the IQ-boosting alleles produce autism instead of a more “classical” low IQ that genetic load alone would have had…

    • melolontha says:

      Has paternal/maternal age been eliminated as cofounding factor from parents IQ/autism correlation? This seems a possible explanation, if high-IQ parents have children significantly later…

      The ‘note’ in the second image mentions adjusting for maternal and paternal age (among other things), and it doesn’t look like the adjustment makes an important difference.

      • gkai says:

        Yes indeed. paternal IQ/ autism seems pretty much decorrelated, which is strange given the 100% confidence that autism genes are IQ boosting.
        Maybe the pattern is something like this (calling “rare” IQ/autism boosting alleles IQb):
        IQb + low genetic load : extremely high intelligence, slight or undetectable autistic tendencies, strongly tenchnically inclined
        no IQb + low genetic load : high intelligence, sociable, not so technically inclined
        IQb – medium genetic load : high intelligence, autistic – highly functional autist or autistic tendencies, strongly tenchnically inclined
        no IQb + mgl: normal
        IQb + high genetic load: autism with lowish IQ, always diagnosed as autist
        no IQb + hgl: very low IQ, non-autistic mental disorders

  11. eterevsky says:

    Is it possible to measure autism on some scale, rather than as a binary category? It would be interesting to see more fine-grained dependency between autism and IQ, i.e. people that are 30% autistic have average IQ X.

  12. Chalid says:

    Is there precedent for this tower-versus-foundation model being useful for any other health issue, or was it just made up specifically for autism?

    • JungianTJ says:

      If a high quantity of a good thing comes with a risk of problems that reduce the quantity again, then you could understand the ability to keep the problems at bay as „foundation of the tower“. (So I don‘t understand why confidence goes down so much for point 5 in the conclusion of the original post, from 80 to 25 percent.)

  13. magnacarta says:

    Each of us have different stories to share.
    I married an autistic woman. She is a university qualified computer systems engineer. One of our children is an Asperger (labelled as high functioning autistic in the DSM). I recognised my son was different at about 15 months old. He taught himself the alphabet before he was 1 year old.
    We are both university qualified. Both moderate to high IQ. We had children at around 30. My wife’s father (an engineer) is likely autistic. I have no family history of autism. I am known as a word smith (and it’s an important part of my career, but I’m not in journalism).

    My son was measured using the full Weschler and most of his scores were 96-99th percentile (that’s VERY high). His processing was below average. I have always considered him more naturally gifted than me.

    Of the autistics I personally know (through common support groups), the Aspergers all have a parent who is an engineer or computer scientist of some sort.

    We have another child who shows no autistic traits. He has the making of a superb engineer.

    Keep in mind, my sample size is small. However, I wouldn’t be surprised if a statistically high proportion of people here are autistic or have autistic family members.

    One other observation that may interest people. We were grouped with autistic including low functioning autistics. If the mums with low functioning autistic children, the ones I spoke to described the fathers as losers or deadbeats. I’m not suggesting this accounts for the high/low intelligence as the sample size was small. I recall it because the lowest functioning children all had loser fathers. The difference between high and low functioning autistic children is significant.

    • chaosmage says:

      Is it kind of like a bimodal distribution, i.e. most kids will be pretty clearly high or low functioning, with little in between?

      What are the percentages of children who fall in each group?

      • CatCube says:

        It could be one of the categorization problems Scott was talking about in “Against Against Autism Cures.” If you’re not “low-functioning” but don’t have strong enough autistic traits to be clearly a high-functioning autistic, you don’t have “autism” you’re “just a little weird?” It seems that a process like this could conceptually produce an artificial bimodal distribution, even if the traits aren’t bimodally distributed. My stats-fu isn’t strong enough to know if this actually pencils out, though.

        Heck, maybe there are two different underlying conditions here that have similar enough symptoms that we categorize them together (per Scott’s observation in his “The Body Keeps the Score” review that we tend to categorize psychiatric disorders heavily by effect, because the cause is so murky).

    • theredsheep says:

      Potential confounder: having a low-functioning autistic child is highly stressful and often unpleasant. Being a deadbeat or loser, or being perceived as same, may be a result, not a cause, of the specifically low-functioning types. For example, they may dislike the father for not doing enough when the kid is self-harming from overstimulation. Or the dad may have left or tuned out because he couldn’t handle the situation (but could have handled a socially inept kid who won’t shut up about WWI airplane models). Their career progress might have stalled due to stress and sleep deprivation. Etc.

    • My son was measured using the full Weschler and most of his scores were 96-99th percentile (that’s VERY high). His processing was below average.

      Many such cases with Aspergers (me too), whereas for high functioning autism the deficit is inverted for processing vs visuo-spatial. The DSM-5 change was a mistake in my view because high functioning autism and AS genuinely seem to be quite different in structure, and it seems as though there were some low level political motivations for folding them together (see upthread). This makes it difficult to analyse high functioning autistics separate from those who would have previously been distinguished as having Aspergers.

      EDIT:
      Or maybe not, because the researchers just act like they are cognitively distinct “subtypes” anyway, so it might not make much difference in practice.

      From this study Scott linked:

      In most GWAS studies there has been little evidence of heterogeneity of association across phenotypic subgroups. In this study, however, we see strong heterogeneity of genetic overlap with other traits when our ASD samples are broken into distinct subsets. In particular, the excess of alleles associated with higher intelligence and educational attainment was only observed in the higher functioning categories (particularly Asperger’s syndrome and individuals without comorbid intellectual disability) – and not in the other/unspecified PDD and intellectual disability categories. This is reminiscent, and logically inverted, from the much greater role of spontaneous mutations in these latter categories, particularly in genes known to have an even larger impact in cohorts ascertained for intellectual disability/developmental delay91. Interestingly, other/unspecified PDD and atypical autism also have a significantly higher PRS for neuroticism than childhood autism and Asperger’s. These different enrichment profiles observed provide evidence for a heterogeneous and qualitatively different genetic architecture between sub-types of ASD, which should inform future studies aiming at identifying etiologies and disease mechanisms in ASD.

      The strong differences in estimated SNP heritability between ASD cases with and without intellectual disability, and highest in Asperger’s provide genetic evidence of longstanding observations. In particular, this aligns well with the observation that de novo variants are more frequently observed in ASD cases with intellectual disability compared to cases without comorbid intellectual disability, that IQ correlates positively with family history of psychiatric disorders92 and that severe intellectual disability (encompassing many syndromes that confer high risk to ASD) show far less heritability than is observed for mild intellectual disability93, intelligence in general94 and ASDs. Thus it is perhaps unsurprising that our data suggests that the contribution of common variants may be more prominent in high-functioning ASD cases such as Asperger’s syndrome.

      Interestingly, this one addresses the P>V, and V>P subtypes and suggests:

      children with ASD did not show the commonly understood characteristic profile on the WISC either in terms of PIQ-VIQ discrepancy or in terms of peak skills on particular WISC subtests.

  14. overdreamer says:

    Isn’t autism a bit more sensitive to increased parental age than many other mental disorders? If it is, that could explain some of these results.

    >They are being positively selected, ie increasing with every generation, presumably because people with the genes are having more children than people without them.

    I think this was also true for some genes supposedly predisposing to ADHD and Schizophrenia. (Although, aren’t gene-behavior connections in general pretty suspect, as per your previous post?) https://slatestarcodex.com/2019/05/07/5-httlpr-a-pointed-review/
    https://sci-hub.tw/10.1176/appi.ajp.2018.18070881

    • Lambert says:

      Can we quantify the causality going the other way?
      i.e. do people predisposed to have autistic kids tend to have them later?

      • hls2003 says:

        I would expect that to be the case based simply on the premises. If smart people are predisposed to have autistic kids (which Scott seems very confident of), another trait of smart people is that they tend to get more education than average. Getting more education is strongly correlated with later childbearing / siring. Based on those two correlations, I would be very surprised if it didn’t add up to a correlation between predisposition for autism and children later in life.

  15. Gurkenglas says:

    If relatives of autists are smarter, then that sounds to me like a regression towards the mean. Instead of autism causing low IQ, low IQ (in a mental architecture that typically causes high IQ) would cause autism, and then of course when you look at the relatives of a downwards outlier you will find higher IQ.

    • Gurkenglas says:

      (When your edit window expires before you submit the edit, it just discards your text and you have to retype it? -.-)

      In the graph of relatives, are autists typically on the border or inside of high-IQ regions?

      Perhaps different people train their brain using different optimization strategies, and when you optimize too well you risk ending up in the autistic basin of configuration space, which does well in the environment a fetus or baby faces.

    • zqed says:

      Regression to the mean is not at all what happens here. According to the linked table, the relatives of autists have higher scores on the Børge Priens Prøve than the general population. Contrast with the relatives of mentally retarded people, whose intelligenxe tends to be higher than that of their retarded relative (due to the effect you describe), but still significantly lower than the population mean.

  16. Garrett says:

    How much do we know about how intelligence develops and what might impact it?

    Probably wrong ideas generated from your writing here:

    We know that core elements of cognition not only do develop early in life, but *have* to develop in life. For example, the experience of people who are blind at birth and then are given sight as an adult vs. those who lost vision as a child and then have their sight restored as an adult. It’s similar with hearing.

    Could there be some sort of input(s)/connection(s) in early development which is responsible for the development or unfolding of intelligence like that? If missing, intelligence develops a lot less/later/incompletely?

    What led me to ask this is stories of people who were deaf from birth who’d get cochlear implants as adults. Because they lacked the development of the auditory processing, it was very draining for them because it mostly came across as a large amount of noise. This meant that they’d routinely turn them off. Now imagine that someone was in that condition but *couldn’t* turn it off. I could easily see someone unable to focus, stuck in a room banging their head just trying to make the noise go away.

  17. alwhite says:

    I don’t have time to go through all the links this morning but wanted to add some more studies.

    Here’s a study that claims “Few adults with autism have intellectual disability; however, autism is more prevalent in this population. Autism measures may miss more women with autism.”

    Another study showing how autistic people have higher prevalence in inpatient psychiatric units.

    And an article by John Elder Robison pointing out that most of our data on autistic people come from children and the extreme cases. We don’t have an understanding of how autism exists in the general populace so our data is skewed towards the worst because of where that data comes from.

    This is the only study I know that has actually measured autism rates in the general public.

    • Aapje says:

      This is the only study I know that has actually measured autism rates in the general public.

      Spoiler alert: ASD in adults was estimated to be 1%

      • alwhite says:

        And this is a productive comment because?

        • antipodeanlurker says:

          … because it is in line with the current prevalence of autism diagnoses in children, but a bit lower. This shows a) there isn’t much of a childhood autism epidemic going on, it’s more likely that previous underdiagnosis occurred. and b) people diagnosed with autism using current criteria are quite prevalent in the community and presumably functioning in their communities.

        • Aapje says:

          @alwhite

          I didn’t understand why you left out the actual finding, which seems to be an important fact to note.

          So I gave that fact to save others the effort to dig for it, while trying to give a subtle hint to you that it is better to include the finding.

  18. Aftagley says:

    These questions interest me because I know a lot of people who are bright nerdy programmers married to other bright nerdy programmers, and sometimes they ask me if their children are at higher risk for autism. While their children are clearly at higher risk for autistic traits, I think they want to know whether they have higher risk for the most severe forms of the syndrome, including intellectual disability and poor functioning. If we take the Ronemus and Gardner studies seriously, the answer seems to be yes. The Gardner study seems to suggest it’s a very weakly elevated risk, maybe only 1.1x or 1.2x relative risk.

    Say you are a bright nerdy tech-aligned person who is attracted to other bright nerdy tech-aligned people; you have a medium preference for eventually having kids but you really don’t want an autistic child. Is there anything you can do in terms of testing or practice to reduce your risk?

    I’m not especially neurologically atypical, but enough that I’m not confident I might not pass down some potentially deleterious genes. This fear is reinforced by the fact that of my 12 cousins, 1 is severely autistic, another one is autistic, but can function, 2 are diagnosed aspie and another 1 is an edge-case. This might sound mean, and I’m very sorry if this offends anyone, but just seeing what my family has gone through, I would definitely find having a severely disabled child worse than not having one and would likely be unhappy with a child any more than slightly aspie. (This sounds harsh, but I want a child at least as normal as I am).

    What are the best practices here?

    • viVI_IViv says:

      What are the best practices here?

      Breed with a Stacy or a Chad. /s

      Seriously, what sort of advice do you expect?

    • hls2003 says:

      Have children earlier, maybe? Increased parental age correlates with autism risk, AIUI. And if you do have a more challenging child, you’ll have more energy to deal with it as a younger parent.

    • antipodeanlurker says:

      Of the 5 relatives, how many are males?
      I would recommend having girls through IVF if that is legally allowed in your jurisdiction?
      And also reproducing younger (banking frozen eggs and/or sperm early for the IVF).

      • Aapje says:

        Girls also get autism, but seem to stay undiagnosed much more often.

        • hls2003 says:

          Perhaps this is an obvious point with an obvious answer, but wouldn’t female prevalence of diagnosed autism be expected to be markedly lower? As far as I can tell, the explosion in diagnoses is primarily driven not by the very low-functioning (everyone could already tell that non-verbal folks had something wrong with them) but by increased diagnosis of medium-to-high-functioning people.

          Since the definition of autism (as noted by Scott and discussed elsewhere in the comments) is primarily symptomatic, that means you only get the diagnosis if you display symptomatic behaviors meeting a certain noticeable threshold. Most of these seem to be related to social ability. If females have a higher average baseline social competence, it would stand to reason that even the impaired girls would remain above the diagnosis threshold in greater numbers. E.g. Being 1.5 standard deviations less socially adept than the average girl might make you weird, but not problematic, while being 1.5 standard deviations less socially adept than the average boy might make you a classroom / societal problem.

      • Aftagley says:

        Of the 5 relatives, how many are males?

        3, but the serious cases are all male.

        And also reproducing younger (banking frozen eggs and/or sperm early for the IVF).

        Does affliction rates vary with age of both parents? I don’t know why, but my prior is that the mother’s age matters more.

        • eric23 says:

          I don’t know. I have a female acquaintance who was diagnosed with autism. At times she has been productively employed in a technological field. But her personal life is a mess. She’s a real handful to deal with and that deters quality men from being interested in her. And her failure to understand social cues lets her be taken advantage of by lower-quality men, who have sexually abused her on multiple occasions. So is she a “non-serious case”? She’s capable of living without a caretaker, but other than that, the impact on her quality of life seems pretty serious.

    • The Nybbler says:

      No guarantees in biology, but you can try to find a future partner with no such family history. Lots of bright nerdy tech-aligned people don’t.

  19. marc200 says:

    Guys — autism is not a thing. Like many DSM diagnoses it is a junkheap of often unrelated symptoms wrapped up in a single label for essentially politically reasons. Coming up with elaborate speculations as to why a unitary “autism” would be characterized by wildly divergent and contradictory symptoms (like being very bright and perceptive and being completely non-verbal and unable to learn) just leads you further down the rabbit hole. It would be better to try for more precision/clarification concerning the form of social or linguistic impairment you are concerned about and investigate that. There is no reason to think that e.g. the kind of social impairment characteristic of bright but shy and social awkward people has much if anything to do at a deep level with the impairment experienced by non-verbal children with profound intellectual disabilities, despite the fact that we have built a social system that applies a similar label to both.

    But don’t take my word for it, check out the excellent 2016 lit review by Waterhouse et al which concludes:

    No unitary ASD brain impairment or replicated unitary model of ASD brain impairment exists. ASD core diagnostic symptoms are not uniquely linked and are only very rarely expressed without nondiagnostic symptoms. ASD has no reliable early predictor, no unitary developmental course, no unitary life outcome, no unitary recurrence risk, no unitary pattern of BAP features, and no standard homogeneous subgroups.

    In sum, these findings suggest that neurodevelopmental social impairment exists in varied forms with varied etiologies and varied pathophysiologies, but the ASD diagnostic criteria do not identify a valid entity

    Link to the entire piece — https://link.springer.com/content/pdf/10.1007/s40489-016-0085-x.pdf

    For all his skepticism in some areas, Scott is remarkably accepting of the overall conceptual framework of psychiatric and psychological diagnosis despite its glaring philosophical flaws, which have been pointed out by many (see e.g. the Council for Evidence-Based Psychiatry in the UK on the lack of validity of diagnostic categories in psychiatry — http://cepuk.org/unrecognised-facts/diagnostic-system-lacks-validity/). It is where his bread is buttered professionally, but I wish he would think more deeply about it.

    • dogiv says:

      Then why is there a genetic link? If the genetic risk factors for high-functioning and low-functioning autism have a lot of overlap, as Scott suggests, then that strongly implies a connection in etiology.

      • marc200 says:

        But I don’t think there is overlap between the genetic factors for high-functioning and low-functioning autism. Low-functioning autism has all kinds of genetic linkages to all kinds of different genetic syndromes that cause brain damage, damage that often extends beyond autistic symptoms. High-functioning autism looks more like the same polygenetic soup one finds with other complex personality characteristics like intelligence or Big Five character traits. Given the numerous different genetic correlates to autism, if one wants to preserve the diagnostic validity of autism by finding a single etiology, one might hypothesize that all of these genes affect a narrow set of brain circuitry in a similar way, but as discussed in the Waterhouse paper people have tried to do this and (per her lit review) failed to do so.

        With that said, I’m not an expert and there is a vast literature rummaging around for genetic correlates to autism, so I’m mostly going by the linked paper above. But in general there will always be a hereditary component to personality characteristics that is probably linked to genes and one will be able to find correlations, but that is a long long way from understanding etiology or validating a single diagnostic label for a broad phenotype.

  20. Maxander says:

    The “Tower vs Foundation” model seems to treat intelligence as if it’s an attribute like height or arm length; the modification of some static variable. But by “intelligence” mean the ability of a complex system (a brain) to operate effectively; so “being smart” is much more similar to “being able to run really fast” or “having a good immune system.”

    So by comparison, I could imagine that there’s a bunch of gene variants that increase limb length, and that if you have a certain number of these you get some nice Usain Bolt proportions that help you win races. If you have too many, your legs are bizarrely long twigs that you totter around on like a newborn giraffe, and you’ve just gone to far so you won’t be a great runner. The physical kinematics of the human body dictate some ideal limb proportions for the 500-meter dash (say), and that’s a target your genetic makeup can miss in either direction.

    The entire idea of a gene “for intelligence” is just as silly as a gene for “running fast.” There are genes which have lower-order effects, and intelligence or running are higher-order capabilities that emerge from the lower-order system in complex ways (indeed, there’s probably several more layers than that.) There’s obviously no reason genetic factors would aggregate in a nice linear fashion in that kind of situation. (And, just to acknowledge what I suspect was the subtext of this entire post, that’s really bad news for people who want to get significant IQ gains from genetic engineering.)

    • marc200 says:

      Thumbs up to this perspective. Comparing intelligence to running fast understates the issue, it is much more like a complex rules-based game such as football, baseball, etc. for which certain genetic inputs might be extremely valuable but there are numerous different ways to arrive at the end point of being good at the game, and numerous different ways that even people genetically gifted in some important capacity can end up being bad at playing the overall game.

    • NovaByblos says:

      Yeah, your reaction is what I was scrolling down here to say, but you beat me to it. 🙂 Your model of “too much of a good thing” seems to, to my non-doctor, non-expert mind, fit perfectly.

      You mentioned “height”, which I think is a good example too – we have studies (possibly not good ones) showing that taller people have better outcomes, get treated nicer, are more attractive – but people who get too tall have bad medical problems, shorter life expectancy.

      • Simon_Jester says:

        Even if intelligence is analogous to “running fast” or “having a good immune system,” the tower/foundation model is still workable.

        If you keep stacking up genes that in isolation would improve your running speed, but that involve trade-offs or may have adverse interactions with each other by both modifying the same chemical pathway or something… Eventually you have too many and things collapse. Each gene’s benefit is based on underlying assumptions about how the body works, and if you change enough of those assumptions, the entire edifice becomes a failure.

        To some extent, incompatible advantages can ‘stack’ by affecting entirely different things (one runner with salutary mutations affecting his legs, his lungs, and his adrenal gland)… But to some extent they can’t.

        Intelligence may be very much the same way.

    • Shion Arita says:

      Yes, I was thinking something very similar: Broadly speaking, it’s pretty clear that intelligence differeces come from differences in brain structure or organization, macroscopic and/or microscopic. The genes that control the development of that definitely don’t do so in a linear or easily explainable way. For example, why do people with down syndrome or williams syndrome have differently shaped faces? On the first level it’s obviously ‘because of the genetic abnormalities they have’, but it’s not like we could look at those gene differences and predict that those features would develop in that way. Or in the case of plants, their leaf structures are generated by a fractal pattern, and even small changes in the parameters of that fractal cause large changes in the ultimate pattern in a way that is chaotic and difficult to predict. And intelligence, being a function of the brain itself rather than directly of the genes that generate it, would in this analogy depend more directly on something more like the shape of the plant’s leaves than the internal fractal parameters that generate it.

      So it may not just be something like ‘too much of a good thing’ but ‘the relationship between the seed parameters (genes) and the result (brain architecture/intelligence) is something that’s chaotic rather than simple.

    • etaphy says:

      My relatively uninformed opinion would be: what the mutations might be affecting in lower-order terms are some subtle parameters of prior mapping, or something like a lower threshold for what’s considered to be sufficient fit for model validity, or what kinds of sensory inputs to give particular significance to. Given that many drugs both medical and recreational appear to affect similarly subtle parameters broadly through relatively simple molecules, there’s no reason to assume they can’t be much more specifically fine-tuned through genetics.

      Maybe specific areas for different such quirks of processing have some optimal Goldilocks zones for their values, which if all fit might give you some sort of the upper bounds of human intelligence-maybe there are multiple such optimal/viable configurations, however hitting all the targets simultaneously must be relatively rare, and some of the targets might be smallish, others might be essentially fixed without some de novo mutation for likely good evolutionary reasons.

      It wouldn’t surprise me if what’s perceived in higher order terms as stereotypically autistic might be a series of misses on such optimal targets, over- or under- shoots, and then you end up in a situation where a child has difficulty working out how to discard pareidolia from their mental model of the world say just because of a specific kind of synapse oversensitive in some particular way, or later they engage in repetitive behavior instead of altering their mental model to accommodate more unpredictable interactions because it appears more efficient to their brain to conserve the model because of the overstated unpredictability of the alternatives, et cetera.
      Some might get lucky and be able to manage the negatives from such unusual prior models away by the effects of other quirks, say having abstract thoughts feel more natural through some particular tweaks making the net their active inference casts ‘wider’ with higher error tolerance or by lowering the interconnectivity between different oversensitive mental modules that they don’t interfere with one another as much or whatnot – then you get high-functioning autistic individuals, maybe Newton’s one cliched example of that sort of configuration.
      The Tower vs Foundation model is probably an oversimplification of something like this.

      My personal bet would be that different sorts of mutations associated with autism could be slotted into abstract prior-mapping-affecting categories manifesting through tiny quirks of cellular chemistry, and some degree of interplay between them and those that affect general nervous tissue metabolism and anatomy might also be the case.
      And the ‘extreme male brain’ hypothesis in that light might just be an artifact of a statistical consequence of GMV in that set of traits, as it could be less reproductively costly to have a less stable architecture for those types of parameters in males due to higher reproductive payoff of hitting one of those ‘optimal configurations’ in terms of number of offspring, as well as lower individual investment.

  21. DinoNerd says:

    This post is missing a discussion of the changing meaning of “autism”. Now that Asperger’s syndrome has been removed from the DSM, in favour of something understood by lay people as “autism”, I’ve become autistic, along with a fair number of other geeks, nerds, etc. (*)

    It’s not hard to measure my IQ, except that it generally falls at the top of the scale where the tests don’t make effective distinctions. (Or did, 20+ years ago when I last took an IQ test.) On that and other evidence, it’s pretty clear the above article isn’t talking about autistics like me. But it’s frankly unclear who it is talking about. Non-verbal people requiring institutionalization? (I.e. the classic stereotype.)

    (*) to complicate matters, in normal usage, I’m “autistic”, but in many usages I wouldn’t be described as “having autism”. Some vaguely similar term like “high functioning autism” would be used. (Note that if autism is really one thing [not a complex syndrome], with high and normal functioning individuals, and high vs normal correlates with IQ, then if you only measure the “normal” autistics, you’d get results like the above, due to sampling bias. No need for (extremely valid) concerns about whether IQ tests accurately measure (normal/low) functioning autistics, e.g. the non-verbal.

    One other effect of this change in labelling, is that studies of “autistics” before and after the renaming can’t be combined into any kind of meta-study, unless they are both much more detailed in their description of the grouyp they are measuring, and happen to target the same people.

    This would be a more useful post if it addressed this change in terminology.

    • marc200 says:

      The Waterhouse study I linked above (https://link.springer.com/content/pdf/10.1007/s40489-016-0085-x.pdf) argues that even before the autism-aspergers lumping made things worse, ASD samples within single studies (let alone meta-analyses) were not useful for scientific inference because the “autism” diagnosis lumps together so many wildly disparate developmental and social issues. As the authors put it:

      Because the ASD diagnosis is defined by just two symptoms from an individual’s complete set of symptoms, an ASD-defined sample is widely over-inclusive, capturing many different lifelong conditions that can arise from a complex combination of multiple genetic and environmental factors (Dawson 2016, para. 1), in which the genetic architecture of ASD seems to be different from one individual to another (Huguet et al. 2016, p. 118), and there are many individually varied connections between brain and behavior (Kennedy et al. 2015, p. 81). In sum, the heterogeneity of risk factors, brain impairments, and non-diagnostic symptoms in an ASD-defined sample blocks valid scientific inference.

  22. Björn says:

    I spent a lot of time reading about giftedness (having a IQ > 130) in the last few months, specifically educating gifted kids and teenagers, and I found out quite some things about how giftedness can interact with social ability:

    It turns out that giftedness can be quite a burden for children and teenagers, as not many of their peers can relate to them and vice versa. This can lead to bullying and social isolation, which in the long term can cause depression and the failure to learn social customs. There is another problem: Giftedness also increases empathy, which in the wrong environment can be a bad thing. Think about it this way: When most other people behave incomprehensible for you because they don’t see the same emotional nuances and act on their impulses, it is hard to learn social customs, even though you theoretically have the potential to be very skilled socially.

    There is yet more. Giftedness makes people more idiosyncratic. It’s hard to pin down why this is the case, but I think it is related to the things above, but also to the fact that gifted people tend to be bad at copying other people instinctively. I think this is connected to gifted people tend to be good at finding creative solutions to things rather than the most obvious ones, which in the case of weird walking or talking styles can look like autistic behavior.

    Reading about this seems a little bit strange at first, but I must say, I teach in the mathematics department of a university, and I observe this everyday in my best students. Every year, we get like 4 or 5 students who are just ready for hardcore abstract mathematics from the get go, and since a few years my university allows them to take masters level lectures as early as they want. I would say that 80% of them carry a fair amount of weirdness with them, which shows in speech impediments, idiosyncratic/shitty clothing styles and obsessions with bizarre topics.

    This could be constructed as autistic behavior, but I don’t buy it. The reason is that interacting with those students always feels the same, it doesn’t matter how weird they seem at first. They all share the same hunger for mathematics, and if you get to know those who seem “normal” on the surface, you find out that they have some interesting hobbies as well. The latter are the Feynman types: Goofballs with a myriad of weird interests (Feynman was into playing the Bongo drums and the short lifed Soviet Republic Tannu Tuva, among other things).

    All this leads me to believe that people who are gifted and who’s weirdness is more obvious are often misdiagnosed as having Asperger autism. Or rather, that the concept of Asperger autism is more related to giftedness, and not to autism. I’m very open to people arguing the contrary, but I must say, in 8 years of doing abstract mathematics, I never met an Asperger-like person where giftedness did not seem to be part of the equation.

    • marc200 says:

      There are so many ways to be weird and out of touch with the people around you, related to very different sets of issues and characteristics.

      Observing matters in my local school, I think that child psychiatry is best understood as a means of managing problems that arise in children’s socialization to the school environment, and clusters of diagnoses align with a big three of socialization difficulties:

      “weird don’t fit in” kids are diagnosed as autistic

      “restless can’t sit still or stop acting up” kids are diagnosed as ADD

      “violent/fighting/tantrum” kids are diagnosed with oppositional disorders

      There are huge practical and financial incentives within our system to diagnose all of these sets of kids with some kind of medicalized disorder, regardless of whether any underlying physical mechanism is understood or present (and sometimes it is, but many different developmental issues can lead you down each of these roads).

      As a practical matter of course these disorder clusters can easily overlap and often do, as seen in the most recent DSM when they allowed co-diagnosis of autism and ADHD for the first time (by popular demand pretty much!). But they offer a way of grouping “types” of problem kids that has caught on. At the individual level things are generally way more complex than these symptom-based diagnoses and the key is to ask *why* the kid isn’t mixing well socially, gets into fights / lashes out, or can’t sit still rather than reify it as a brain dysfunction.

      • Björn says:

        I find it absolutely bizarre that the intelligence of kids who don’t fit in is often ignored. It is known that gifted kids can be misdiagnosed as autism, ADHD or both, but many psychologist never bother doing IQ tests, or do them, but then ignore the results.

    • Thegnskald says:

      Slightly differently: Being gifted makes it harder to take certain things seriously. What is the point of arbitrary social rituals?

      Well, there is a point, but it takes a long time to grasp it, at which point you are significantly behind people who went along with them without questioning them. More, you have probably already settled into an existence that minimizes their importance to you by the time you have enough information to piece together the importance of predictability.

      More, by that point your identity probably includes significant aspects which make the idea of becoming normal and predictable less personally palatable.

      • marc200 says:

        Social rituals are incredibly important though, humans are social animals. Being extroverted and popular is one of the biggest predictors of success and longevity, probably more important than IQ frankly although I don’t have the studies in front of me.

        • Alex M says:

          The only reason social rituals are important is because other people say that they’re important, and historically it was important for high-intelligence people to be able to get along with low intelligence people or they would be outcast from the tribe and die. Hence the need to go along with pointless social rituals.

          However, history doesn’t always predict the future. If you get a high intelligence person who is smart enough to create a game-changing technology – for example, militarized AI or super-forecasting technology – then they don’t need to get along with low-intelligence people: instead, low-intelligence people need to get along with them. In a situation where being outcast from the tribe means that the tribe is going to die instead of the outcast, then the tribe had better learn to put up or shut up.

          • Hypoborean says:

            So far, this is still 70% wish-casting to 30% reality-aligned.

            There are a lot more pathways to leveraging raw intelligence to acquire economic status now than there used to be, but very few that get you all the way to the point that you’re fully safe from the tribe (since the tribe is now the nation and is thus also much stronger now).

            Note that billionaires in America are still vulnerable to (using a topical example) wealth taxes, and abroad billionaires in Russia / China must play the “cuddle up to Putin / Xi or die / get put in a camp” game.

            Also, if we do get to a bold new future of hyper-intelligent unilaterism, would the hyper-intelligent even want to sign up for it? Maintaining or constructing rigorous social boundaries would be the only thing that would protect a human with IQ300 from a human with IQ400 (or a seed AI).

          • Alex M says:

            I think you’re exaggerating unilateralism, Hypoborean. You don’t need to be an absolute dictator or god-king to be able to ignore social norms – all you need is lots of money. Just look at Kanye West, for example.

            If I personally were one of these hypothetical genius scientists in possession of technology that could reshape the world, I think that the most optimal strategy for me would be to offer this technology to the U.S. in exchange for a large sum of money – say, $50 million. That would give me more than enough money to give the middle finger to all the ignorant normies in our politically correct society, with plenty left over to hire some bodyguards and build mechanical defense systems to insulate me from any of the outraged pearl-clutching idiots who decided to use physical violence against me when their threat of social sanctioning wasn’t enough. Plus the military is always interested in easy ways to enforce security and they aren’t too bothered by political correctness, so any automated defense units that I might hypothetically build could also be a profitable side-business for me. I’d cheerfully get in bed with the military-industrial complex -in fact, I’d probably be rock-hard the whole time. The military is where the fun toys are all at!

            If that didn’t work because the moralizing pearl-clutching normies insisted on creating arbitrary obstacles in my path to financial success, I would definitely see cuddling up to Putin as a good alternative. It’s much easier to please one person than an entire society, and unlike our judgmental mouth-breathers in the States, Putin at least seems like a pretty pragmatic and reasonable guy. I honestly wouldn’t be surprised if he was a rationalist himself. As long as I kept useful inventions flowing out of my lab, I think he’d pay me well and insulate me from social criticism. Plus being allied to Putin would allow me to strike back at the assholes in the U.S. who treated me unfairly because they were offended by my pragmatic philosophy and refusal to play along with their ignorant social customs, so that would be a nice little side bonus. Revenge can be fun – just ask Taylor Swift!

            All this is a purely hypothetical thought-experiment, of course.

    • It turns out that giftedness can be quite a burden for children and teenagers, as not many of their peers can relate to them and vice versa.

      One solution is to view your family rather than your schoolmates as your peer group. Judith Harris discusses this in one of her books.

      More generally, to relate to those people around you who are also bright.

      • Björn says:

        Yes, of course. I view it as absolutely crucial that bright kids get to socialize with kids like them, and also that someone teaches them how to be a functional person. I get to see what happens to those kids otherwise, they become adults who ace every math exam, but are unable to buy pants.

      • Rachael says:

        Gifted children don’t always have gifted families, though. I was a bit of an outlier in my family, and didn’t really find a peer group I fitted in with until university.

    • albatross11 says:

      n=1, but it was striking to watch my very smart middle kid go from being a near-outcast in his good Catholic school to a guy with lots of friends at his magnet school. It really made a difference for him to get to a school where he fit with the other kids.

  23. quitelikelyblog says:

    Interesting stuff.

    My first thought when hearing that autism intelligence data is that people are much more likely to end up with the “autistic” label if they’re disabled. So the technically autistic community could well be more intelligent on average than the general population, with the people who are actually labeled autistic being a particularly low-functioning subset of that population.

    • Thegnskald says:

      My original post talked about dual axes, functionality and intelligence, with non-functional high intelligence becoming, effectively, a distinct kind of low intelligence. After further consideration, I’m unaware of any phenomenon which would correspond to a low-function, low-intelligence version.

      I’m tending towards a resolution-processing power version now, as I have considered it more, with high resolution being a burden if you don’t have the processing power to deal with it, but my thoughts are incomplete.

    • Rachael says:

      Yes, agreed.

      The Charman et al study finds a correlation between autism and low IQ, by starting with a sample of autistic people and looking at their IQs. That seems like a poor methodology; I think it would be better to start with an unbiased sample of the general population and test them for both autism and IQ.

      Scott mentions selection bias, with people diagnosed as autistic being more likely to be the low-functioning ones, but then he seems fairly dismissive of its importance.

      I’m reminded of the Simon Baron-Cohen theory about autism being like an “extreme male brain” in some sense. (Is that still seen as credible? I haven’t been keeping up.) It is often said that male IQ has the same mean as female IQ but greater variance, so you find more men in both tails of the IQ distribution. Perhaps something similar but more extreme is true of autism: maybe autistic people are more highly represented at both ends of the IQ distribution. (Which isn’t the same as saying it’s bimodal; the distribution still peaks in the middle, but it’s wider and flatter.)

      • Simon_Jester says:

        I’m skeptical of the ‘extreme male brain’ hypothesis for autism.

        Why? Gender dysphoria. My understanding is that gender dysphoria is associated with certain structures or traits of the brain- transwomen tend to have certain identifiable feminized brain structures, and transmen to have masculinized brain structures.

        While autism (or at least diagnosed autism) is rarer in girls, it’s very much not unheard of. If the ‘extreme male brain’ hypothesis is true, then those girls have unusually masculinized brains. If so, we’d expect a much higher rate of female-to-male transgender identity among people on the autism spectrum who, ah, lack that Y chromosome.

        Similarly, you would expect lower rates of male-to-female transgender identity among the population of people on the autism spectrum who came equipped with that Y chromosome. Because a masculinizing influence on the brain would militate against the development of brain structures telling the autistic person “you are actually a girl.”

        Is this result observed?

        I mean, it COULD be that autism somehow exaggerates male brain characteristics in a bunch of different respects, but then doesn’t touch the parts of the brain that affect gender identity. But that wouldn’t be the way to bet, IMO.

        • The Nybbler says:

          Why would you doubt it if you don’t have the data?

          Anyway, if the data were there, I would suggest an alternate explanation — that a significant portion of gender dysphoria is socially mediated, and the autistic are isolated from the social phenomenon. But such an explanation is premature.

        • DinoNerd says:

          FWIW, anecdata in the autistic communit suggests that autistic people are more likely than average to be genderqueer in some fashion, but there’s no anecdata correlating this with chromosomal sex – i.e. no one’s saying this only affects autistic people identified female at birth.

          • Le Maistre Chat says:

            FWIW, I’ve only ever heard biological females apply the term “genderqueer” to themselves.

          • Nornagest says:

            I’ve met a couple AMAB people identifying as genderqueer, but at least an order of magnitude more AFAB ones.

        • Aapje says:

          @Simon_Jester

          Quite a few trans women seem to be extremely masculine in their interests. So I doubt that trans feelings require an opposite-sex brain.

          That said, I too doubt the ‘extreme male brain’ hypothesis in a general sense, although it may be true in a more limited sense (where autism consists of more masculine features for certain parts of the brain, but not all of them).

  24. JASSCC says:

    How does age of parents figure in? I have read before that age of parents increases risk of autism.

    Perhaps what’s happening is that the genetic factors passed along are responsible for generally supporting IQ, but high IQ parents are having children later in life for some reason (more years of schooling than average, early stage demanding careers, saving a nest egg before starting a family). So the genes are a correlation but not a cause of the autism, which is due to factors incidentally related to the high IQ, for example the advanced age of the parents?

  25. rodan32 says:

    I’ll share some info that might or might be helpful, since it’s just an anecdote. I have 5 children. Two of these are diagnosed as autistic, though with very different “forms”. My oldest child, a boy, is very high functioning. High IQ, excellent at math, maybe a little socially awkward but not too bad (we live in a reasonably large town, so there are lots of like-minded kids he can socialize with.) I’d say his IQ is similar to or higher than mine (around 135.)

    My 3rd child, a girl, is much lower functioning. Her condition presents with some speech and language processing difficulties. She’s unable to work in a “normal” public school classroom, but luckily we have a charter school that specializes in children with autism that has worked well for her. Her IQ has only been tested once, by a neuropsych, and came in around 90. That seems too low, based on my interactions with her: she’s not so much “slow” as “odd”, so it may very well be that the test isn’t a good measure.

    A couple of things about her birth stand out when comparing the two with the other children. First, these two children were the only children where my wife was working while she was pregnant. I hesitate to make anything out of this, but whenever there’s a study that links pre-natal stress to autism, my ears perk up a little. Second, my 3rd daughter’s birth was challenging. She was born with the cord around her neck, and we had respiratory specialists on hand to assist. No ICU or anything that severe, but still something that differentiates her from the other children.

    As for genetics, well, my IQ is ~135, and my wife’s is right there with me (~125). My wife’s father would probably be diagnosed as autistic if they’d looked for it when he was younger; very smart guy, but very odd. My wife is dyslexic; I’m not. Our ages when our son (1st child) was born were 27 (me) and 26 (my wife), and for our daughter (3rd child) we were 31 and 30. Our 2nd, 4th, and 5th children are not autistic, so parent age doesn’t seem to have played a role in our case (though we were both relatively young even for our fifth, at 36 and 35.)

    Being able to observe a large variety of children with autism (due to activities at the charter school), I agree that “autism” is a very strange catch-all with many different types of children diagnosed. There’s a lot of work left to do to sort out what being autistic actually means.

  26. Thomas Jorgensen says:

    This is just the cystic fibrosis thing again, isnt it? Genes which are helpful until you have too many of them, at which point you are screwed.

    … I am getting a vision of the future. It is very, very smart, but nobody dares reproduce naturally anymore, not because the equipment doesnt work, but because genetic engineering means everyone pushed every single one of these levers to the setting marked “beyond this point, there be dragons”, and at least half their kids would die or come out completely non-functional if they tried it.

  27. FlorianDietz says:

    As an AI researcher, this actually seems pretty intuitive for me.

    The brain has hyperparameters, just like an artificial neural network. Genes encode changes to these hyperparameters.

    Let’s say for example that parameter A is the learning rate in the part of the brain that deals with reasoning. It’s normally at 5.0 in an average human. The optimal value is at 10.0 (for example). Any gene that increases A in a normal human makes that human more intelligent by pushing A closer to 10. However, if too many such genes coincide in the same individual, then A exceeds its optimum of 10. For some parameters, exceeding the optimum can lead to strange and non-obvious errors.

    For example: As the learning rate increases, a network becomes able to learn new information more quickly. However, once it exceeds a certain threshold, the network starts learning things that don’t make any sense, and then when it tries to correct those mistakes it goes too far in the other direction again. The result is a total mess.

    So you end up with a system where every gene that increases hyperparameter A makes the individual smarter, except that the individual becomes mentally ill if the hyperparameter exceeds a certain threshold. That’s exactly what we observe here.

    I would hypothesize based on this that autistic people usually have parents who are very good at exactly those areas that they have trouble in. This correlation may not be very high though, because genes often code for a lot of different things at the same time so the mental illness may affect a different hyperparameter than the one that makes the parents smart.

    • Scott Alexander says:

      My problem with this kind of model is that some people are geniuses but don’t have autism. If it’s just “intelligence is good up to some point, then you get autism”, how do you explain intelligence sometimes going way past that point without autism?

      • Hypoborean says:

        I mean, doesn’t that just mean there are lots of hyper-parameters and you can push one set outside the optimal range without maxing out everything?

        Very simple and wrong toy model:
        There are 5 hyperparameters, the optimal value for all is 9, and 10+ in any causes autism, and each hyperparameter is driven by many genes and different people have different mixes of genes.

        Score:
        9 6 5 4 3 marries 9 2 5 4 3, children high chance of 10+ on first and autistic, children that are not autistic are just barely on the bright side of normal.

        Score:
        6 6 6 6 6 marries 6 6 6 6 6, decent chance of getting a 9 9 8 7 9 stone cold genius

        Non-autistic geniuses are compatible with the proposed hyperparameter model, unless I’m missing something.

      • albatross11 says:

        Are there other mental disorders that track with intelligence? Some of the genetic diseases of Eastern Eurpoean Jews are correlated with intelligence, IIRC, but I don’t think they have mental effects.

    • viVI_IViv says:

      Let’s say for example that parameter A is the learning rate in the part of the brain that deals with reasoning. It’s normally at 5.0 in an average human. The optimal value is at 10.0 (for example). Any gene that increases A in a normal human makes that human more intelligent by pushing A closer to 10. However, if too many such genes coincide in the same individual, then A exceeds its optimum of 10. For some parameters, exceeding the optimum can lead to strange and non-obvious errors.

      This requires a strongly non-linear sensitivity to hyperparameters, which is unusual for the fitness w.r.t. scalar traits in biological systems, which is usually approx. linear or quadratic around the average and possibly tapers off sub-linearly on the tails. This is more or less a requirement if the system is to be robust and quickly adaptable to environmental variation. I believe this is also the case for artificial neural networks, unless you are in pathological regimes that are too unstable to be useful. If e.g. 10.0 is the optimal learning rate, and 9.5 is good, then 10.5 is typically also good.

      I don’t think you can model the autism-IQ correlation using a simple additive genetic model.

  28. LadyJane says:

    I’m not an expert on biology or genetics or neuropsychology by any means, but I feel like you’re onto something with the tower-and-foundation model.

    I knew a wrestler back in college, really huge and muscular dude. He used to claim that he had an advantage in the ring because his muscle was “all-natural,” which made him stronger than people who got equally buff from working out. This always seemed nonsensical to me, because muscle is muscle; if two people are equally large, it shouldn’t matter whether they were born that way or got that way through exercise. And, being a fairly intelligent person himself, he agreed that it didn’t make much sense, but it didn’t change that fact that he was able to overpower equally muscular people through sheer force in the ring. But one of his trainers actually had an explanation: A lot of people who casually work out don’t exercise all of their muscles equally, so even if they appear to be as strong as him, there could be some muscles that aren’t as developed as others. And in a sport like wrestling, which is heavily reliant on locks, holds, and pins, having those weak points could easily allow someone to overpower you. (Always Sunny parodied this phenomenon with Mac, who only works out his “glamour muscles.”)

    The same principle can be applied to genetic traits. A lot of dog breeds, particularly those that are very small (like Teacup Yorkies), very large (like Great Danes), or otherwise very different from the default canine somatoform (like Pugs), tend to suffer from health problems at a vastly higher rate than mutts or more natural breeds (e.g. German Shepherds, Huskies). In general, the less a dog resembles a wolf, the more likely it is to have musculo-skeletal, respiratory, and circulatory problems. Why is this the case? It’s not because simply being small or being large or having a flat face are inherently unhealthy traits; there are other species that are much smaller than Yorkies, much larger than Great Danes, and have flatter faces than Pugs, without suffering any problems as a result. It’s because we selectively bred dogs to have these traits at a much faster rate than evolution could naturally develop them. As a result, they have ‘unbalanced’ physiologies; their bone and cartilage may not be strong enough to support their frames, their hearts may be too large or too small for their bodies, their nasal passages may be truncated because their snouts are shorter than the rest of their genetic blueprint expects.

    There could be something similar going on with autism. Humans have been selectively breeding themselves since the dawn of civilization, even if we weren’t consciously aware of it; reproductively speaking, the traits that are selected for in an organized society are bound to differ than those that would’ve been selected for in the wild. As a result, we could have similar imbalances. And since mental traits like intelligence, social aptitude, and self-control are the ones most likely to be selected for in society, it makes sense that our imbalances would primarily be mental. Physically, people with autism might have certain brain structures or modes of neural connection that are associated with high intelligence, but without the surrounding neurological infrastructure necessary to support those structures/functions – like the poor Great Dane whose heart is too small and bones are too weak for his frame.

    • albatross11 says:

      +1

      Very interesting idea.

      My vague mental model is that there are a bunch of different parameters that have to be within a certain range to get good performance from your brain. The population mean for many of those parameters is off from the optimal value for thinking, perhaps because there’s some tradeoff encoded there that doesn’t matter as much now as in the past. (For example, a too-big head would kill mother and baby until modern times; now it just means mom gets a C-section.)

      For each parameters, there are many alleles in different locations that can bump the parameters a certain direction–make the baby’s head[1] a little bigger, for example. And so those are genes that show up as giving you a bump in your IQ in populations genetic studies. But if you pile on enough of the genes that increase IQ by increasing your head circumference, you will get a kid with such a huge head that his neck won’t support it and his heart and lungs can’t keep his giant brain in oxygenated blood, and he’ll die as a baby. There’s a range for this parameter, selection in the past has left most of the population too low, but if you push the parameter up too high things will break.

      [1] I assume most such genes do something way more complicated and hard to measure, but head circumference makes a nice example trait.

    • Clutzy says:

      This has nothing to do with the OP, but I will definitely comment on your wrestling example, as a wrestler.

      Its 100% true. When I was in high school one of my coaches called them, “air muscles.” Its when a guy can bench 300, squat 500, etc and still get overpowered by a guy who does half that. Its not about purely not working out, although I would say natural strength is a part of it (almost every wrestler can increase his in-match power through lifting still), its about the muscles being coordinated and firing at the right angles and right times to actually accomplish something. This is why, as a freshman at 125 lbs, I could beat seniors at 160, 171, 189, and 215 lbs on our team, despite all of them crushing me at the bench press. However, I couldn’t beat the Sophomore at 130 who was my workout partner in the weight room and only had 5-10 lbs on me in most lifts. Because he was as skilled as me, and had “real muscles”.

      • Cliff says:

        The reality is that weight training only increases strength in the motion of the lifts that you do. So for example, bench is largely worthless if you’re not pushing people away from you. Big muscles can also reduce range of motion.

        • acymetric says:

          I would file this under “partly true”. Yes, the biggest gains from any given exercise are to performing that specific exercise, but gaining strength that way still allows you to do other mostly unrelated motions that use the same muscle group that you couldn’t have done without those exercises.

  29. Glenn says:

    I think it’s possible that some relatively simple model will be found to explain this relationship, but I also think it’s possible that the mechanics of intelligence are simply complex enough that the explanation is something inaccessible with our current level of understanding. That there is some knob which, as you turn it up, it increases intelligence, until you get it too high and then it blows up due to some fact about the dynamics of the system that we don’t understand yet.

    An analogy: Imagine you have a sound system with a microphone, speakers, and a mixing board. Over the vast majority of the range of the main volume knob on the mixer (assuming the system is set up reasonably), more will be better. Even when you get it high enough, to the point of being “too much of a good thing”, if it’s a rock show some people are still going to want it higher. But there is a critical point in the system, the feedback point, after which more isn’t just “too much of a good thing” — the behavior of the system changes qualitatively and the whole thing breaks down. If you get the volume knob too close to the feedback point, the sound will start to behave strangely — it will “ring” weirdly in a way it’s never done before. If you get much past that, all intelligible sound will cease and be replaced by a terrible noise that wasn’t exhibited _anywhere_ prior in the range of the knob.

    There are ways to reduce the loop gain of the system to increase the maximum “gain before feedback”, but ultimately they involve careful sound engineering — there’s no general parameter you can turn down to fix the problem, because the primary parameter you _want_ is the same one that causes the breakdown. You can squeeze out a little more gain by filtering just the frequencies that feed back, but if you keep going up it won’t take long at all before there are too many to filter. Above a certain level of gain, the system _wants_ to break down.

  30. Hypoborean says:

    Ok, I just got to the end of the comments and no-one else has brought this up yet:

    Isn’t the finding that
    (1) genes that promote autism are correlated with IQ and
    (2) genes that promote autism are being selected for
    HIGHLY surprising?

    I thought we were starting to get clear evidence that genes associated with IQ (through the proxy of years of education completed) were being selected AGAINST in 20th century society, which is intuitive (educated people on average have fewer children, because time during child-bearing years is consumed with schooling).

    Am I wrong about the second finding?

    If I’m not, that suggests one of three things:
    (1) Genes associated with autism aren’t associated with IQ
    (2) Genes associated with autism aren’t being selected for
    or, most spicily:
    (3) Genes that promote autism and IQ are being selected for DESPITE other genes that promote IQ being selected AGAINST, so we would expect to see the representation of autism in (the dwindling fraction of) high-IQ families explode in the coming decades.

    • The selection findings are across historical time, not recent industrial time, and they’ve found that IQ is being selected for as well.

    • LadyJane says:

      Am I wrong about the second finding?

      Yes. To the best of my knowledge, the whole Idiocracy concept (humans are selecting against intelligence because smart people have fewer children) is basically nothing more than a just-so story made up by concerned pseudo-intellectuals. I haven’t seen any real evidence for it, or even any real support for the idea as a serious theory.

      Also, if anything, it would be self-control that’s being selected against by modern reproductive behaviors, not intelligence. The two may or may not be correlated (I’m honestly not sure), but they’re not the same thing. I know plenty of highly intelligent people who nonetheless have very poor impulse control – particularly when it comes to libidinous desires! – and they seem just as likely to have unplanned children as less intelligent people with equally bad impulse control.

      • Hypoborean says:

        Let me see if I can dig up one of the studies that I saw mentioned.

        “There’s a new paper out on how the frequency of variants that affect educational achievement (which also affect IQ) have been changing over time in Iceland. Naturally, things are getting worse.
        We don’t have all those variants identified yet, but from the fraction we do know and the rate of change, they estimate that genetic potential for IQ is dropping about 0.30 point per decade – 3 points per century, about a point a generation. In Iceland.”

        https://www.pnas.org/content/early/2017/01/10/1612113114.full

      • marc200 says:

        you haven’t seen any support for Idiocracy? The damn thing has basically come true! Trump = President Camacho, the contemporary media is just as idiotic as the media in that movie, etc. Drink your Brawndo!

      • I know plenty of highly intelligent people who nonetheless have very poor impulse control – particularly when it comes to libidinous desires! – and they seem just as likely to have unplanned children as less intelligent people with equally bad impulse control.

        Anecdote is not data, which overwhelmingly show a negative correlation between IQ and fertility.

      • Cliff says:

        IQ dropping over recent decades in many developed countries, combined with lower TFR for high-IQ isn’t evidence?

        • LadyJane says:

          According to the most recent findings on the subject, it’s not “many developed countries,” it’s only Scandinavia. In other countries – including those in developed areas like Western Europe and North America – IQ has been on the rise.

          https://www.bloomberg.com/opinion/articles/2018-03-06/humanity-keeps-getting-smarter

          • Hypoborean says:

            Flynn effect vs genetic selection. Overall scores could continue to rise due to the Flynn effect even as intelligence is being selected against on a genetic level, which is what the evidence is suggesting.

            If that’s the case, if the Flynn effect ever slows down or goes into reverse we’d be in quite a bit of trouble, and Scandanavia could be the canary in the coalmine for that process.

            Other interesting question if the Flynn effect is more important than genetics and our mean IQ is 15 points higher than 100 years ago: where are all the geniuses? If the curve stayed exactly the same and there was a correlation between IQ > arb. threshold and massive achievement, massive achievement should have exploded in recent decades.

          • mtl1882 says:

            Other interesting question if the Flynn effect is more important than genetics and our mean IQ is 15 points higher than 100 years ago: where are all the geniuses?

            The mean could easily shift without creating a bunch of geniuses–it is probably much easier to raise the IQ of average or above average people, just through some relatively basic improvements and guidance. Additionally, look at the lives of geniuses throughout American history. I don’t think there was much correlation at all with years of formal education, and the paths their lives and careers took are almost unimaginable now. I would argue the same “standardization” that has raised IQ generally actually de-selects for geniuses in many cases–we spend a lot of time getting people to think and act a certain way and behave “efficiently” etc. It could be that geniuses are less visible with more obscure achievements because they work behind-the-scenes at institutions, whether universities or corporations. That’s assuming that we do have a “missing” group of geniuses–I’m not sure how we track these things, or who you are counting as a genius.

          • dionisos says:

            where are all the geniuses?

            I think, everywhere.
            Even if intelligence where a little lower we would have a lot more genius given the increase in population.

            I am convinced there are a lot more genius now, but :
            – We have limited time to dedicate to know these genius, and a big part of it is dedicated to old ones.
            – Doing some [fundamental] discovery is harder and harder.
            – The top level of any domains is so hard that most people can’t judge anything about it. Some good story make much more in our appreciation of who is a genius than anything else.

          • Cliff says:

            Research suggests that there is an ongoing reversed Flynn effect, i.e. a decline in IQ scores, in Norway, Denmark, Australia, Britain, the Netherlands, Sweden, Finland, France and German-speaking countries,[4] a development which appears to have started in the 1990s

            the Flynn effect has ended for the majority in developed nations

            Research that has examined whether g factor and IQ gains from the Flynn effect are related have found there is a negative correlation between the two

          • @Cliff
            If the g factor and IQ gains from the Flynn Effect are negatively correlated, then does that mean that IQ tests are gradually being more and more poorly designed over time?

          • albatross11 says:

            Hypoborean:

            Other interesting question if the Flynn effect is more important than genetics and our mean IQ is 15 points higher than 100 years ago: where are all the geniuses? If the curve stayed exactly the same and there was a correlation between IQ > arb. threshold and massive achievement, massive achievement should have exploded in recent decades.

            I think (as an interested amateur, not an expert) that the most plausible answer is that human intelligence depends on getting lots of stuff right. The genes and developmental stuff is one part, an enriching environment in childhood is another, and the physical requirements (adequate nutrition and good health and few toxins present during critical parts of brain development in childhood, continued good health in adulthood). Take any part away, and you lose some intellectual ability.

            In this model, there’s a ceiling to your intellectual ability that’s set by each of those components. The people who have the genes to be the smartest humans can hit somewhere close to their potential by happening to have the right environment, enough food and no development-stunting diseases during critical times, etc.

            This is consistent with the existence of people in the past who seem to be about as bright as humans get. I’m not a great student of philosophy, but I don’t think most modern philosophers look at the writings of Plato or Descartes or Kant or Hume and think they were intellectual lightweights. There’s probably not anyone working in math right now who is any smarter than Gauss was. For that matter, some amazing ingenuity went into ancient engineering–enough so that I think we still don’t know exactly how the Incas built Machu Pichu or the Egyptians built the pyramids. A reasonable guess there is that the smartest engineer on the pyramid project was not much dumber than the smartest engineer working on the Apollo project–they came up with brilliant one-off solutions to their problems, which didn’t stick around because there wasn’t much use for them in day-to-day building problems.

            The Flynn effect seems to be about raising the floor by building proper sewers and water supplies, eradicating parasites, vaccinating everyone against common childhood diseases, making sure nobody starves and also adding nutritional supplements to foods so brain development isn’t stunted by deficiency diseases, eliminating heavy metals from the environment of small children, making sure everyone goes to school, etc. We can’t do much about the genetic or random developmental stuff, but we can make sure most of the things that have stunted peoples’ intelligence over time go away.

            That should give us more geniuses, but not a shift in the distribution where now we get a Gauss/von Neumann/Newton/Kant/Einstein born in every town. The high end should stay the same, in much the way that the high end of human lifespan is only a little longer now than in 1800–it’s just that a lot more people make it to 90 or 100 with modern medical care.

            This can coexist with a loss in the number of people with the highest potential intelligence–if we’re building sewers and safe water supplies at the same time we’re convincing our smartest 5% of people to take a vow of celibacy, we’ll get that pattern. Or if (as really happens) it turns out that smart, ambitious people usually have their kids a decade later than dumb, unambitious people and usually the smart, ambitious people only have enough time in their fertility window for one kid while the dumb ones have time for three.

  31. marshwiggle says:

    Epistemic status: wild speculation backed up only by anecdotal evidence and theory.

    I have a possible explanation for the observed distribution of intelligence in autistic people. I am one of those people who has to run emulators to appear socially normal. I also need alternative means to handle a wide variety of learning and perceptual tasks. Some of the ‘software’ I am running to handle those tasks ‘manually’ requires huge amounts of processing power, resolution, intelligence, and or ability to abstract. I also have more ability to perceive what processes I am running. That leads me to the following guess.

    What if autism does two things: it has a tendency to break enough ‘normal’ processes to make someone rather nonfunctional, and it also has tendencies to increase perceptual ability and intelligence. There is then a threshold effect – with enough perceptual ability and intelligence, the developing brain finds ways to route around the broken ‘normal’ processes to get stuff done. That, combined with intentional ‘software’ creation can lead to more or less high functioning autists. Fall short of that threshold and you have a pile of severely broken processes. This model predicts something I believe is true – many low functioning autistic people have real talents lurking in there, it is just really hard for them or others to access those talents because of all the broken stuff.

    This fits with my observations of my family (parents, siblings, kids) and other science majors in college. I have many fewer data points on low functioning autistic people, who I have mostly met through church. I am not particularly confident in the theory either in its parts or in whole.

    • antipodeanlurker says:

      n=1 description of autism and speculation on etiology.
      I am a highly educated (PhD) male parent of a son with autism (aged 17 so in high school), who is in a Special Education class, so I fit the phenotype described here. My parents were optometrists. My wife also has a university degree, her father was an engineer and his 3 siblings had careers in operations research, research scientist, school principal. There are no other cases of autism diagnosed in the immediate family.
      I believe autism is largely genetic and the genes that are associated with autism mainly perform functional roles in brain development and brain function. Autism is associated with caesarian section https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2749054
      (my son was a Caesarian delivery) but that was because his head (=brain?) was too big to go through the birth canal!
      We often joke that my son learned English as a second language, it appeared that his primary language acquisition pathway just wasn’t there. His auditory channel is very slow and only turns on when you say his name or repeat the question. He visually learned letter and number orders before turning 2. He initially appeared to learn the names of items when we attached labels with printed words on them around our house.

      I favour the Maxander description of the genetic defect in autism, too much brain development can be bad.
      There is also some link to male sex, I do believe that there are higher levels of males with low intelligence autism.

      Tongue in cheek (Swift “Modest Proposal”) style interventions given this model are:
      autism polygenic risk scores attached to Tinder profiles
      Engineer/scientist social mixers with barbers/hairdressers and fashion merchandisers
      Titrated use of alcohol in pregnancy and general precautions regarding healthy eating during pregnancy

      • albatross11 says:

        If autism is a result of piling up too many high-IQ alleles, then you’d expect to see far more of it in the children of scientists, lawyers, engineers, doctors, mathematicians, etc., than in the children of janitors, file clerks, policemen, nurses, etc.

      • albatross11 says:

        Two of our three kids had extremely large heads as babies. (All three were C-section births.) None are autistic, but both boys were a little socially awkward–basically a couple years behind their classmates in this area. This seemed to balance out by late middle-school/early high-school.

  32. deciusbrutus says:

    >The Gardner study seems to suggest it’s a very weakly elevated risk, maybe only 1.1x or 1.2x relative risk. But the Gardner study also ceilings off at 90th percentile intelligence, so at this point I’m not sure what to tell these people.

    You tell them that the study that tried to address that question suggests that they are between one point one and one point two times as likely; in other words, at least 83-90% of their chance to have autistic children is baseline.

    • Scott Alexander says:

      Suppose the only study on age-related mortality shows that people older than 30 are 1.2x more likely to die in a given year than people younger than 30, but does not give any other information.

      A 120 year old comes into my office and asks what her risk of death next year is.

      Do I say “1.2x that of someone under 30?” Or do I say, “Not enough data to answer, plausibly extremely high”?

      • deciusbrutus says:

        You tell them that the study that tried to address that question suggests that they are between one point one and one point two times as likely; in other words, at least 83-90% of their chance to die in any given year is baseline.

        The fact that nobody has noticed and studied an effect in a subgroup is evidence that the subgroup does not have a large effect.

  33. deciusbrutus says:

    Why conflate “diagnosed with autism” with “has autism”?

    People with average or above average intelligence and autism are much, much less likely to be diagnosed, because one contributing cause of a diagnosis is seeing a psychiatrist, which is much more likely for people who are having trouble in classes.

    Smart people who can manage things by themselves and mask the effects of their autism from others don’t get diagnosed, because among other things they don’t want to see a psychiatrist and can figure out how to stop people from forcing them to; they can also know the ‘correct’ answer to psychiatric interviews and can figure out that the things currently bothering them are less likely to continue if they provide the correct answers.

    Pretending that people with autism diagnoses are typical of people with autism is useful when trying to treat the symptoms of autism; but ignoring less symptomatic people results in ineffective studies when trying to figure out causes and effects.

    Figuring out the causes of neurotypicality in intelligent people would go a long way, but it’s hard even to measure how neurotypical a given smart person is.

  34. theodidactus says:

    My wife is autistic. I am not, but we both initially bonded over (I don’t know a way to put this delicately or more exactly) being objectively very “smart” people that don’t “perform” intelligence very well, albeit different ways. We both spent a lifetime getting scolded because we were obviously too smart for the dumb stuff we kept doing (grades and the like). When people came to our defense, it was generally with an idea I didn’t like too much: that maybe there were “different types of intelligence” and if we sucked at [x] we must surely be good at [y]…like life was an RPG character generator or something. I never liked that.

    The tower model looks close to my personal model, which has absolutely no experimental testing besides how the two of us feel about it…but generally I call it the S’more model. If we’re gonna sit down and make S’mores, we need marshmallows, chocolate, graham crackers, and a crackling fire. Four things that make one delicious combination.

    …but here’s the thing, I can’t come back from the store with six bags of graham crackers and confidently say “alright, we’re gonna make a BUNCH of s’mores now.” To really make a lot, we need all marshmallows, chocolate, AND graham crackers…and adding a bunch of one factor does absolutely nothing, and may in fact hinder the whole operation…similarly, it’s fallacious to suggest that because some factors dial up, they must all be “good things” that should be dialed up without limit…that crackling fire is a necessary component, but it doesn’t need to “go any higher”…if it does, if we have a big ‘ole bonfire for example, the whole process actually gets harder.

    Broadly, intelligence might be, say: pattern recognition, ability to readjust priors, visual imagination, factual recall, procedural memory etc. These things are incredibly synergistic in some people, allowing them to do exceptional things, we notice this…so we start to associate them with one another…but the presence of one means absolutely nothing about the presence of any others in general, and you need to “scale up” all or at least many of them to get any good returns…worse, some factors may not need to “go any higher”, otherwise they’ll just be a hindrance.

  35. wanterp says:

    The dangling paren is disquieting enough to have prompted me to make an account:

    most of the other statistics mentioned here).

  36. elspeth diana says:

    i do wish you wouldn’t use the phrase “autism risk”.

  37. Jakub Łopuszański says:

    Under https://slatestarcodex.com/2019/05/07/5-httlpr-a-pointed-review/ Murphy wrote:

    3: genetic games of chicken.

    disorders like huntingtons where it’s theorized that there’s a tradeoff between risk of death and IQ. too many CAG repeats and the area destabilizes and you get runaway anticipation that wipes out later generations of your family.

    https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6013750/

    Increasing repeat length was associated with higher GAI scores up until roughly 40–41 repeats

    But >39 repeats means risk of the disease.

    Epistemic status: not too bad, certainly good enough to have my old head of department, a grand old neurogeneticist interested in this if still maintaining healthy scientific skepticism.

  38. AlexSpark says:

    I think the study chosen to show low IQ is really common in autism is a really weird one, and probably a sampling issue.
    It’s a small sample (156) and it’s 10-14 year olds, not adults. Autistic people have weird developmental trajectories.
    Compare to this study https://bmjopen.bmj.com/content/9/8/e029040.full based on the scottish census data. Sample: literally the whole of scotland. And that finds that only 15% have intellectual disabilities. That’s high, compared to the general population (below 1%) but 50% is weirdly high.
    Given the problems measuring IQ in autistic people (spiky profiles, huge differences between ravens vs. weschler) it’s also possible that intellectual disabilities are overreported at 15%. Autistic people being diagnosed with intellectual disabilities before they’re able to to talk and not really being “undiagnosed” when they’ve demonstrated higher levels of ability is very much a thing.

    • babarganesh says:

      yes, that study seemed really bad.

      also didn’t say anything about recruitment. how could they recruit in such a way as to get a representative sample? my kid who is on the spectrum is 12 and ok functioning (doing well in middle school) – i can’t imagine what would get us into the therapist office to do an IQ test. when we did a comprehensive neuropsych his IQ came back above average (range 115-125ish depending on how you weight scales) which seems about right.

      also, i just read through the swedish study cited in the post. it separates ASD, ADHD, and ID (intellectual disorder). unsurprisingly, ID is inversely correlated with IQ; ADHD also shows a strong inverse correlation with IQ. however, ASD without these two confounders shows no correlation with IQ.

  39. nimim.k.m. says:

    I have not read the other comments, so maybe someone already pointed this out, but I wish to file a complaint. The post is laden with unsaid assumptions that do not make sense to me.

    Question the first. Why on Earth one would assume that the effect of genes that contribute to IQ should be linearly additive and increasing to ad infinitum? Optimizing a complicated biological machinery thing to do whatever intelligence is most likely a very difficult thing for that machine to do. It sounds very plausible that the effect of gene variant a and the effect gene variant b alone tweak the machinery in a way that increases its capabilities, but when applied together, they could have a catastrophic collision. (Apologies for a machine metaphor that is certainly misleading, but better than anything else than I can think of.)

    You don’t need a tower-and-foundation model to explain how both higher than usual prevalence of intelligence-associated genes in autistic individuals and extremely intelligent non-autistic people. One just needs to consider a model where a brain is a complicated object whose organization and function depends on changes in the genetic blueprint in non-monotonic ways, and it is likely that all combinations (attainable via sexual reproduction) of genes that are beneficial alone or in some particular arrangements are not even more beneficial together; some of the potential arrangements are bound to be non-optimal or even catastrophic. Maybe even only a small fraction of potential arrangements are just right to produce ever-increasing levels of some capability in the phenotype.

    Let me write toy model, where intelligence Q is a function f of all heritable biological parameters that could possibly affect brain function (let us say, all possible variants of human genes in a DNA sequence of certain base pair length), denoted by g_1, …, g_n. g_i = 1 if the parameter is present, g_i = 0 if not. This is very very crude model that does not take into account how DNA and sexual reproduction works: it is middle of the night and that stuff is not my strong suit; however, I like this description and with my limited understanding, it may even carry some resemblance to reality:

    I will bet a considerable sum of money that a model where one simply assumes that Q = f(g_1, … g_n) = a_1 g_1 + … + a_n g_n for unknown constants a_i does not describe how these parameters affect intelligence (“get all g_i that are positively correlated with IQ, combine them, and you get a super-enlightened being!). It stands to reason that f can be very intractable, and resolving that function is the difficult question where there is science to be done.

    Question the second. Why on Earth people take a decomposition of intelligence quotient such as the v-p-r, while itself fairly reasonable sounding but also very broad, and take seriously attempts to shoehorn that as the biological mechanism behind a very handwavy metaphor-like explanation such as “tower and foundation”?It is blatantly obvious that v-p-r is a very crude model that should not make sense. 3% confidence is far too large.

    • mtl1882 says:

      One just needs to consider a model where a brain is a complicated object whose organization and function depends on changes in the genetic blueprint in non-monotonic ways, and it is likely that all combinations (attainable via sexual reproduction) of genes that are beneficial alone or in some particular arrangements are not even more beneficial together; some of the potential arrangements are bound to be non-optimal or even catastrophic. Maybe even only a small fraction of potential arrangements are just right to produce ever-increasing levels of some capability in the phenotype.

      Agree with your post, and especially with this part. Intelligence is clearly not a neat and predictable thing, at least beyond certain broad generalizations or thresholds. And many types of successful arrangements/balances exist–brilliant minds can look very different, and a trait or dynamic good in one person can take another person down if the overall arrangement is out of whack–actually, the same trait can lead to wildly different outcomes in the same person. Virtually everything has a failure mode. Intelligence seems much more similar to personality than height in terms of complexity and diversity of manifestations.

    • quanta413 says:

      Question the first. Why on Earth one would assume that the effect of genes that contribute to IQ should be linearly additive and increasing to ad infinitum? Optimizing a complicated biological machinery thing to do whatever intelligence is most likely a very difficult thing for that machine to do. It sounds very plausible that the effect of gene variant a and the effect gene variant b alone tweak the machinery in a way that increases its capabilities, but when applied together, they could have a catastrophic collision. (Apologies for a machine metaphor that is certainly misleading, but better than anything else than I can think of.)

      Dunno, but you don’t need to know the exact structure of the genotype to phenotype map to check whether the assumption is empirically correct. Knowing how much of the genetic variance in intelligence (say IQ for an easy measure) is additive is probably good enough to figure out how off the purely linear assumption is. My vague recollection on this is a large portion of population variance in IQ is additive (like 30 or 40%?), but the non-additive portion is of roughly comparable size.

      You could compare those estimates for IQ to those for height. The additive proportion for height is likely higher. Lazy googling yields claims around 80% narrow-sense (additive) heritability for height which is comparable to the total heritability of IQ.

      That’s sounds like a significant amount of epistasis in IQ, so I agree it’s not that weird that a few people end up very different than you’d expect from a purely additive genetics + additive environment model.

    • Why on Earth one would assume that the effect of genes that contribute to IQ should be linearly additive and increasing to ad infinitum

      It’s not an assumption, it’s a result:

      https://www.nature.com/articles/ng.3285

  40. thetitaniumdragon says:

    I think an important possibility to consider is that some or all of the correlation we’re observing is simply due to bias in the data we’re using, rather than any actual relationship between high IQ and autism.

    One big category would be selection bias due to diagnosis bias.

    For example: people who are smart are more likely to seek out a diagnosis for their kid who has something wrong with them. Moreover, if your family isn’t too bright, you might not even recognize that there’s something funny about your latest kid. With more aggressive screening for such things, we would see a much smaller diagnosis gap.

    There’s some evidence for this – between 2002 and 2017, we saw a decrease in the diagnosis gap between rich and poor white people, but the gap between rich and poor Hispanics and rich and poor Blacks remained constant. This would be consistent with the idea that more aggressive screening (which is likely to be most aggressive at schools full of white kids because those schools have better resources for such things on average) would cause this effect size to shrink considerably if not vanish.

    https://www.spectrumnews.org/news/race-class-contribute-disparities-autism-diagnoses/

    Another, potentially related, possibility is that we are disproportionately likely to diagnose white kids with autism rather than other forms of mental disability. Black and Hispanic children are less likely to be diagnosed with autism than white kids, but a lot of black, Hispanic, and Native American kids have various forms of learning disabilities. Indeed, blacks, Hispanics, and Native Americans all have higher levels of overall learning disabilities than whites do. Because there’s a correlation between IQ and race, overdiagnosing the children of a higher IQ group with autism (or underdiagnosing the children of a lower IQ group with autism) results in a “correlation” between IQ and autism which is actually just caused by racial bias in diagnosis rates. This could also tie back in to socioeconomic status – better off parents may be more likely to have kids diagnosed as having autism instead of some other less “acceptable” mental disability, which could also potentially explain the racial disparities. This would also explain why there’s a larger gap in autism diagnoses between rich and poor blacks and Hispanics than rich and poor whites – more affluent blacks and Hispanics have access to the same resources as affluent whites do, while poor ones have worse resources than poor whites.

    And of course there’s always survivor bias. For example, in World War I, we saw an increase in “head injuries” after we started giving soldiers helmets – but the reason wasn’t that soldiers wearing helmets were more likely to get injured, but because previously, the injuries to their head simply would have killed them and they would have been marked as dead, not as having head injuries. Because they survived due to wearing helmets, and we were only measuring *surviving* soldiers, helmets appeared to increase the number of head injuries, rather than decrease their severity.

    And we know that there’s differential infant mortality rates. Blacks have an infant mortality rate more than twice as high as whites do. Poor people have a higher infant mortality rate than rich people do. And rich people, on average, receive better medical care – particularly rich white people. Higher IQ people have lower infant mortality rates than lower IQ people. And autistic people have more health issues than non-autistic people do. One possible explanation is simply that autistic children born to poor parents/less intelligent parents/disadvantaged minorities are simply more likely to die in infancy/early childhood and thus never be diagnosed as being autistic to begin with. Improvements in health care thus cause an “increase” in the number of autistic children born to more affluent families, but what’s actually happening is that we’re just seeing more of them *survive*.

    Indeed, higher survival rates over time might also explain why we’re seeing a real increase in the number of autistic children over time – more of them are living long enough to be diagnosed with autism in the first place. This would also neatly explain how a primarily genetic disease is becoming more prevalent – people were always producing autistic children, they just weren’t surviving before.

    There’s one other possible source of systematic bias in the data: living and available parents, especially fathers.

    Low IQ people are more likely to end up dead or in jail, and are less likely to be in stable nuclear families. They also have higher rates of having children with multiple partners.

    If you’re trying to compare the IQs of parents and children, you need to have both the parents and children, and be sure that the parents are in fact the parents of the children in question. Low IQ fathers are less likely to be available due to death/imprisonment/not being around anymore/maybe not even knowing who the father is.

    Thus, any sample of fathers is likely to overestimate the average IQ of fathers on average because fathers are around to be measured (aren’t in jail, aren’t dead, didn’t just take off) have higher average IQ than those who who aren’t around to be measured.

  41. benf says:

    Probably not the most important feature of this research but it’s a thunderous argument against modern day IQ-obsessed cryptoeugenicists. Genes are not something where you can say, this is good, more of it will be better.

    • Genes are not something where you can say, this is good, more of it will be better.

      If this is good, then by definition, more at the margin will be better. There’s a question if you want to go really far out, outside the modern human range, and you might get unforeseen side effects. But the basic eugenics of moving intelligence higher at the margin through differential fertility, or the soon-to-be possible embryo selection for intelligence, this is not an argument against that.

    • deciusbrutus says:

      Indeed. More copies of a good gene are almost universally detrimental.

  42. DragonMilk says:

    “required to answer an examiner’s quetsions[sic]”

  43. dark orchid says:

    First impression on reading this, apart from “excellent post, thank you!”: it sounds like it could be something where each contributing factor individually increases intelligence a bit, but if you go over a certain limit then it tips over and you end up with lower intelligence again. Possible metaphor: overclocking your computer a bit makes it better (faster), but push it too far and you start getting weird random crashes – and end up getting less work done overall.

  44. NooneOfConsequence says:

    This is a great post. I’ve looked into this topic myself, not professionally or anything-just as an interested lay researcher with a science/statistics background- and had a less well-articulated formulation of the tower vs. foundation model in mind for awhile, too. I’ve since backed off of it a bit, maybe not because I think its wrong but because there’s possibly a larger and more interesting explanation for what we see, which is that both the collection of deficits that go with low-functioning ASD and the neural efficiencies that go with high IQ are both at least partially caused by a twitchier immune system, which in turn is creating a different profile of synaptic pruning as the brain develops.

    High IQ’s appear to be selected for, as does ASD; while the the survival advantage of the first is obvious, the second is a really high cost to pay from a fitness standpoint, and I’m skeptical that the IQ on its own would be worth it as a survival advantage. But you know what really has provided a massive survival advantage over the last few millenia of people clumping up together in cities? Immune systems that greet pathogens with a really forceful response on a hair trigger. And so humans recently had/still are having a massive selection for genes that require lower thresholds for immune responses, and genes that “outsource” some of the immune de-escalation responsibilities to gut microbes and parasites, etc in order to get a faster jump have proliferated to the point where a lot of these are “normal” now; this is the hygiene hypothesis or “old friends hypothesis” and the best outline of I can offer for it is the book “An Epidemic of Absence” by Moises Velazquez-Manoff.

    So, one can see how this would obviously lead to an increase in autoimmunity, but in one of the later chapters AEoA, Velazquez-Manoff also shows how autism (a least a lot of the time) is an inflammatory condition: for instance, autopsies uniformly showing neural inflammation, people with ASD’s are more likely to develop autoimmune diseases than the general population and are more likely to have family members with autoimmune diseases, and – for a few subjects – treatment with immune-modulating intestinal worms providing some symptom relief even. Maybe this is old news to most people, so I won’t belabor that point any more, but what about the link between high IQ and a different immune function? Karpinksi et al. surveyed ~4000 MENSA members and found that in addition to being at a signficant increase in risk not just in ASDs and mood disorders but also in food allergies, asthma and 2x RR of autoimmune conditions compared to the general population.

    It makes sense to me that the microglia that are involved in both the elimination and growth of synapses could end up producing some brains that are less functional and some that are more functional vs. the “wild type” based on how the extra pruning is done, what parts of the brain it’s done in, and when it’s done. And thus a little more involvement from the microglia in utero or during early childhood might lead to a more efficient neural configuration and a higher IQ, especially if the involvement is more concentrated in the brain regions with more of an evolutionarily enforced margin toward resilience/survival rather than cognitive performance in normal humans. Too much difference from wild type synaptic pruning and you’ve got pathology in addition to the possible, but not certain, optimization.

  45. Strawman says:

    Pure speculation on my part, but I wonder if part of the link between autism and high intelligence mightn’t be explained by a non-linear relationship akin to that between arousal and performance. Plausibly, a certain amount of narrowness and rigidity of focus will improve cognitive performance, whereas too much will cause impairment by making it too hard to get the big picture, switch attention and deal with competing stimuli etc.
    Superficially, this seems more plausible to me than the tower vs. foundation model, especially since it’s type of phenomenon we already observe for better-understood things the brain does, e.g. with stimulant use (arguably, it also explains a number of non-clinical nerdy traits rather nicely).

  46. wiserd says:

    I wonder what the genetics for this would look like. A simple model might be something like the genes for color vision. A mutation in the pigment for the red cone allows a person to see ‘two shades of red.’ Having more diverse cones means that they can differentiate shades of red that other people would see as identical. However being homozygous for this new mutation would lead to a person being red-green colorblind. The mothers of red-green colorblind individuals are frequently tetrachromats.

    If we can isolate these genes for high intelligence theoretically we could duplicate them so that a person might be more certain to have both a normal gene and also a variant gene. Kindof like we could genetically engineer all people to be tetrachromats.

  47. ArawnGraalrd says:

    I was under the impression, that IQ was culturally biased, as DNA Markers would be.

Leave a Reply