There are two kinds of people in this world: those who are deeply suspicious of twin studies, and those who welcome their confirmation that that their identity is fixed and heritable. I’m in the first group. I always have been. Maybe it’s something in my genes.
I first encountered the popular accounts of the Minnesota twin studies when I was a teenager, seeing the scientist and some of the twins doing the rounds of the afternoon talk shows — I think I saw them on the Mike Douglas Show (I’ve dated myself now). I remember them going on and on about the amazing similarities between the twins who had been raised apart. They both married women with the same name! They drank the same brand of beer! They were both volunteer firemen! They gave their dogs the same name! But while there were some recognizable similarities in the pairs, at the same time the obsession with superficial trivia wrecked the credibility of the stories. What? You’re trying to argue that my pet’s names are somehow encoded in my genome? It seemed to me that what we were seeing is echoes of similar culture in their upbringing (later confirmed: most of the twins weren’t really ‘separated’, but were raised by different relatives).
I also saw psychological tropes that ought to have been recognized. These were people who were rewarded for finding coincidences, and they avidly complied, and the scientists were readily accepting of coincidences as evidence of fundamental causal similarity. I was exposed to this pop genetics at the same time I was reading Fate magazine with a critical eye, and the stories were similar. I’d see stories that claimed to confirm the fact of reincarnation, for instance, by compiling lists of similarities between the contemporary claimant and their past life incarnation. They have the same birthday! Note the resemblances in this old-timey photograph! He lived in the Civil War era, now he is a Civil War re-enactor! He died in a fire, and now he’s afraid of fire!
It was exactly the same. That bugged me. And to this day I still see people touting the old twin studies as conclusively demonstrating the genetic basis of personality and intelligence, declaring that it has been positively confirmed that the heritability (a word they often don’t understand — genetically, it has a very narrow and precise meaning that isn’t exactly what they think it is) of intelligence is exactly 50%, meaning that half your IQ is determined by your genes (again, that’s not what it means), and therefore we should be more concerned with breeding intelligent people than teaching people. I also see this fandom coupled with other ugly associations — racists love it, as do Libertarians and simple-minded techno-fetishists. There are definitely genetic contributions to brain development and behavior, but human twin studies are deeply flawed and prone to exaggeration.
Stephen Hsu is a member of the gullible second group. He has posted a reply to my criticisms of his claim that we can readily ramp up human intelligence to reach an IQ of 1000 because hey, intelligence is obviously heritable. The twin studies say so.
1. Cognitive ability is highly heritable. At least half the variance is genetic in origin.
Yes. Cognitive ability is highly heritable, in a general sense — it’s built into the nature of being human that we all have a certain kind and degree of intelligence. We’re people, not chimpanzees or cats, and our brains are part of our genetic and evolutionary heritage. It is also the case that there is some variable component, the heritable part in the strict genetical sense of the word, and some of this variance is, as he says, genetic in origin. I’d argue that there are good reasons to doubt that it is as high as he claims. I’d also suggest no matter what the specific number attached to this variance might be, it does not imply that you can cheerfully fine-tune the genetics to ramp up intelligence to any arbitrary value you can name.
Talk to me when you’ve clearly defined “intelligence” (other than as the ability to perform well on intelligence tests) and have figured out how to isolate it from all the other variables, like class and education, that it is entangled in. Inventing new labels (like “cognitive ability”) does not increase understanding. Lumping a complex network of interacting traits into a single category of “intelligence” actively reduces understanding.
Much of Hsu’s complaint consists of incoherent pronouncements about the number of factors that contribute to “intelligence”. It’s thousands of alleles! It’s thousands of single-nucleotide polymorphisms! It’s tens to hundreds of genes!
Here’s the bottom line: I don’t care. Nobody knows yet, and it doesn’t matter whether it’s one gene that directly regulates “intelligence” or tens of thousands that modulate it, since you can’t even define genetically what is being measured. I suspect that there are going to be a lot of factors underlying the development and function of the brain, everything from cardiovascular health and nutrient processing and immune system effectiveness to genetic variants that affect synaptic vesicle fusion rates. Everything is pleiotropic. We’re not going to be able to sort out all the components cleanly, and in particular, we aren’t going to be able to genetically engineer a gene that has a positive contribution to intelligence without mucking up some other attribute of the organism.
But Hsu has an answer to that. Unfortunately, he only has one answer, because he seems to have read only one paper in genetics, and so he repeats it over and over.
It’s the damn chickens.
He showed them in the Nautilus article that first got my attention.
He used them in an article from last summer.
He drags them out in his rebuttal to me, which is almost entirely a copy-pasted repost of that earlier article.
Jebus, dude. Read another paper. The effectiveness of selection for quantitative traits is well known, with many examples. Look up milk production in dairy cattle. Bushels of corn produced per acre. Egg production by chickens, if you must indulge your avian fetish. Racehorse performance over time…oh, wait. No. Maybe you shouldn’t look that one up.
Thoroughbred horses have been bred exclusively for racing in England since Tudor times and thoroughbred horse racing is now practised in over 40 countries and involves more than half-a-million horses worldwide. The genetic origins of the thoroughbred go back largely to horses imported from the Middle East and North Africa to England in the late seventeenth and early eighteenth centuries. Since the establishment of the Stud Book in 1791, the population has been effectively closed to outside sources, and over 80% of the thoroughbred population’s gene pool derives from 31 known ancestors from this early period. Despite intense directional selection, especially on the male side, and the generally high heritabilities of various measures of racing performance, winning times of classic races have not improved in recent decades. One possible explanation for this is that additive genetic variance in performance may have been exhausted in the face of strong selection.
That paper actually finds that there is plenty of genetic variance present in the population of horses. It’s just not translating into rapid performance increases. It’s almost as if, when dealing with complex behavioral traits with many contributing factors, like running, that simply selecting for a single property, like muscle mass or mammary gland output or egg production, doesn’t work.
Hmm. I wonder if intelligence might not involve many interacting factors?
The IQ fanatics all seem to be stuck in a simplistic state of mind. They have a number, meaningless as it is, and now they must increase the number, as if it were grams of chicken meat, or dps in World of Warcraft, or number of tickets won at skee ball. And they think, since they have a number, it must be easy to make it a bigger number, just by doing more, and that there can be no limit to that number.
Anyone who suggests that there might be limits or that there might be more to the game than single-minded pursuit of a single reward is an obstructionist who will be proven false, eventually, by inevitable destiny.
So I’ve been getting this kind of nonsense from the usual techno-optimists.
— TJ Bradders (@AMilitantAgnost) April 5, 2016
PZ sounds like the guy that said humans were incapable of traveling 100 MPH.
I don’t know “the guy”, and I presume it was some old-timey dude talking about traveling on trains or whatever, but we’re talking about biology and genetics. I think I can confidently say that no human will be able to run at 100mph, ever, despite the fact that the men’s mile record has been reduced by 17% in the last 150 years. Running performance is a number! The number is getting better! Therefore, we’ll eventually have people running at 100mph, right?
The fastest land animal can run 70mph for short distances. That’s an upper bound for a specialized organism. There actually are physical limits to performance, and there are almost certainly physical limits to the performance of the brain, as well. We just don’t know what they are, yet. But blithely proposing that there are no limits, and that an IQ of 1000 is achievable, is as ludicrous as proposing that we tweak human genetics to get people outrunning cheetahs.
Furthermore, that paper on racing has an interesting idea: it compares the rates of improvement of running performance between horses, subject to intense breeding programs to increase racetrack speed, and humans, who are not being intentionally bred for running. Humans have been improving faster than horses, in spite of not benefitting from genetic optimization. How can that be?
For man, almost certainly the huge technological advances, increased awareness of physiological change during exercise and novel training methods have played a greater part than genetic endowment (Macarthur & North 2005). Since the 1950’s interval training, that is training at different paces during a single session, has become an integral part of all athletes programs and has undoubtedly improved fitness and thus racing times (Kubukeli et al. 2002). There is also a psychological incentive for human athletes to not only win races but to win them in record-breaking times (Abbiss & Laursen 2005).
It concludes that it would be interesting to see if “adoption of modern training methods to racehorse training, applying the principles of exercise physiology, nutrition and interval training” might produce more gains in horserace performance.
At a guess, intelligence is probably more similar to running speed than body mass, requiring multiple parameters to improve in concordance, and with all kinds of nested interactions. Intelligence isn’t something you can just bulk up with more protein!
There are a lot of other issues Hsu simply ignores. Two of them are evolution and ethics — he seems completely oblivious to both.
What are the evolutionary concerns? I brought them up before, but let’s simplify the problem here. The average IQ is 100 (by design). Yet there are plenty of people with an IQ of 150. If there are clear advantages to human beings getting smarter, shouldn’t we see some evidence that that positive variation increases its representation over time? But we don’t. We aren’t smarter than, say, the people who lived in the Renaissance era, or in classical Rome. Why don’t we look back on people in those earlier periods as, perhaps, all mildly retarded? I have a couple of explanations for why smarter than average people haven’t taken over the world.
Selection doesn’t see an advantage to an IQ of 150. Sorry, smart people. Other people with an IQ of 100 seem to be well-adjusted, successful, and happy, and can cope with the modern world just fine. They also have children. There’s no particularly good reason to pressure the population to increase the frequency of hypothetical intelligence alleles.
Higher IQs may actually be a detriment. Maybe there’s some side-effect of increased intelligence that selection can see — like delayed onset of reproduction, or an increase in the frequency of asocial cluelessness. Hsu’s superbright IQ 1000 humans might die out in a spiral of self-possessed misanthropic obsession.
There may simply be physiological limits to how big an ape’s brain can be — it eats 20% of our energy output already — and genetics is approaching that limit. Maybe the reason we haven’t seen a big change in the capability of the human brain over the last 100,000 years is that we hit the ceiling then, and human genetic variation is just bouncing us around that level. We can’t exceed it without a radical alteration in the structure of the nervous system or our physiology, which won’t be achieved by Hsu’s proposal of just optimizing existing variants.
The variation isn’t primarily a consequence of heritable traits at all — it’s a result of complex interactions between genes and environment, and is most readily affected by external influences, the equivalent of “principles of exercise physiology, nutrition and interval training” for the brain. So if there is some great benefit to having an IQ of 150, in the population at large you’ll find people who meet that criterion despite have detrimental SNPs, because they had the benefit of good nutrition and education, and people who have the SNPs Hsu wants to promote but don’t meet the criterion because they lacked good nutrition and education.
Maybe, just maybe, he ought to consider the fact that human health and happiness and success aren’t a consequence of a single quantitative trait. Maybe we’re complicated with many roles and many ways of meeting life’s challenges. Maybe we aren’t battery chickens.
But possibly the biggest flaw in his whole scheme is that even if he were right about the genetics, he hasn’t put the slightest thought into the ethics, or even simply the how of carrying out his grand scheme. He’s talking about experimentally manipulating an incredible number of human zygotes, with no appreciation of failure rates or even the possibility of side effects, and raising them to adulthood. He wants to manipulate the genetics of the nervous system, and doesn’t seem to be aware of how fraught with risks that is — and ethical experimenters always should consider how to handle failed outcomes.
Let’s imagine some hypotheticals.
So you’re tinkering with genes that affect brain size. Sometimes you’ll mis-edit, and you’ll get microcephalic embryos — easy to handle, just abort early. But sometimes you’ll succeed and get children with the larger brains you want, and then you discover that the cardiovascular system needed a tuneup, too. You’ve got a barracks full of 5 year olds who have constant debilitating migraines. Or they’ve got big brains starved for oxygen that are unable to develop. What do you do?
You’ve found some genes that directly affect the sensitivity of synaptic transmission. You’ve edited your flask full of ova (by the way, where did you get them?) to include this feature. They develop just fine. The nursery is full of babies. Every time you turn the lights on they go into massive epileptic seizures. You can’t even raise them to the point where they can start learning stuff because they’re suffering so much. What do you do?
The wealthy racist moneybags who is funding all your work is horrified to discover that you’ve obtained most of your gene variants from African populations (because that’s where most human genetic diversity is found) and cuts off your funding until you promise to only use pure Aryan sources. What do you do?
You’ve got a roomful of eager, healthy young students who’ve all had their genes tweaked in various ways. You’ve got to rank them. Which one is best? What do you measure?
You’ve got a roomful of eager, healthy young students who’ve all had their genes tweaked in various ways. It turns out that human children do not spontaneously aquire “intelligence” by instinct. Instead, they’ve got to be educated. Suddenly, your variables have spiraled out of control! Different children — excuse me, different polymorphisms — respond in different ways to different educational methods! What do you do?
These are just a few examples of considerations that are essential to contemplate before you blithely suggest that you’ll just use CRISPR/Cas to edit a bunch of genes in human embryos to create an imaginary genetic superman. Hsu does not or cannot consider them. He reminds me of the traditional physicist joke, which is where I’ll end this.
And that’s what I think of Stephen Hsu’s “genetics” calculations.