You are currently viewing High heritability does mean that social interventions likely won’t work well

High heritability does mean that social interventions likely won’t work well

The scientific literature is full of claims like this one (Paige Harden 2021):

This hereditarian pessimism about the possibility of social change, however, is based on a fundamental misunderstanding of the relationship between genetic causes and environmental interventions. As the economist Art Goldberger quipped in the late 1970s, your genetics caused your poor eyesight, but your eyeglasses still work just fine.3 That is, eyeglasses don’t just help with the environmentally caused portion of bad eyesight. They help with all of your eyesight, regardless of whether it is genetically or environmentally caused. In so doing, they serve as an outside intervention that severs the association between one’s myopia genes and having functional vision.

And Neven Sesardić (2005) compiled a bunch of quotes:

PKU is one hundred percent hard-wired in the genes. Yet it can be effectively cured with a one hundred percent environmental intervention. (Collins et al. 2001)

it is well known that heritability does not imply immalleability or fixity. A characteristic can possess a heritability of 100 percent and yet be fully remediable by environmental interventions. Perhaps the best known illustration of this point is phenylketonuria (PKU). (Lilienfeld & Waldman 2000)

PKU is a genetically determined, recessive condition that arises due to a mutation in a single gene on chromosome 12 (with a heritability of 1), and yet its effects are highly modifiable. (Sternberg & Grigorenko 1999: 541)

PKU is a trait with a heritability of 1.0. But its expression can be drastically altered by a change in environment. PKU thus demonstrates that biology is not destiny. (Paul 1998: 180) A related misconception is that the effects of a genetically determined abnormality cannot be changed by environmental manipulation. Again, this is wrong even with single gene diseases. The example of the inher- ited metabolic disorder phenylketonuria (PKU) illustrates this point well. (Rutter & Plomin 1997: 210)

Heritability does not imply inevitability, because environment can deter- mine the relative impact of genetic variation (GE interaction). For example, phenylketonuria – a genetic cause of mental retardation – is 100% heritable, yet affected individuals can avoid its consequences by eliminating phenylalanine from their diet. (Gray & Thompson 2004: 477)

Even Plomin’s behavioral genetics textbook (Knopik et al, 2018, 7th edition):

A related point is that heritability does not imply genetic determinism. Just because a trait shows genetic influences does not mean that nothing can be done to change it. Environmental change is possible even for ­single-​­gene disorders. For exam- ple, when PKU was found to be a ­single-​­gene cause of intellectual disability, it was not treated by means of eugenic (breeding) intervention or genetic engineering. An environmental intervention was successful in bypassing the genetic problem of high blood levels of phenylalanine: Administer a diet low in phenylalanine. This important environmental intervention was made possible by recognition of the genetic basis for this type of intellectual disability.

For behavioral disorders and dimensions, the links between specific genes and behavior are weaker because behavioral traits are generally influenced by genes and environmental factors. For this reason, genetic influences on behavior involve probabilistic propensities rather than predetermined programming. In other words, the complexity of most behavioral systems means that genes are not destiny. Although specific genes that contribute to complex disorders such as ­late-​ ­onset Alzheimer disease are beginning to be identified, these genes only represent genetic risk factors in that they increase the probability of occurrence of the disorder but do not guarantee that the disorder will occur. An important corollary of the point that heritability does not imply genetic determinism is that heritability does not ­constrain environmental interventions such as psychotherapy.

I submit that these quotes are somewhere between technically correct but misleading, and plain incorrect. Let’s take it case by case, using intelligence as our example. Let’s not fool ourselves, most people discuss this stuff use eyesight and PKU as examples, but really what they are trying to argue is that high heritabilities are compatible with typical social science inspired interventions. They are not.

Our best estimates of intelligence show the heritability is about 80% for adults in Western countries in modern times. Of the remaining 20%, some of this is fake variation caused by imperfect tests (measurement error). Of the remaining, a large part is inherent biological randomness that we cannot alter. Of the remaining of the remaining, let’s assume that about 5% is due to the factors that vary between homes that people grow up in (the shared environment, or C²). What does this tell us about the outcome of any specific intervention? The answer is that it depends. Let’s imagine that our intervention to increase intelligence is giving people books for their homes. We know that households vary in the number of books at home, and this has been a talking point for sociologists for a couple of decades. One can still find this idea in e.g. PISA reports (2018, volume 2):

One of the home possessions that most clearly distinguishes students of different socio-economic status is the number of books at home. While 46% of advantaged students reported having more than 200 books at home, on average, this is the case for only 6% of their disadvantaged peers. Advantaged students also reported a greater availability of other educational resources, such as educational software. In addition, more than 90% of advantaged students but only 69% of disadvantaged students, on average across OECD countries, reported having a quiet place to study at home and a computer that they can use for schoolwork.

But we also know that there are public libraries, and books are cheap, so any child that actually wanted to read something could easily acquire a physical book. With the internet, it is not difficult to find ebook pirate sites that will give you pretty much any book. So for all intents and purposes, access to books has already been roughly equalized and we didn’t see a big closing of the intelligence gaps between homes. Clearly then, books at home are not causal for intelligence. Rather, the number of books at home is caused by parental intelligence and their academic interest, and these traits are heritable, so that’s why we see the correlation to the intelligence of the child later on. So clearly, any intervention designed to improve intelligence by distributing free books would not work.

Now, think about other things that vary between homes. Some homes have parents that read to their children more than others. Some have many computers, laptops, tablets etc., and some have few. Ownership of these appliances have become very widespread in the last decade and we haven’t seen any big reduction in intelligence gaps between families either. So these weren’t really important either. They were just correlates, but non-causal for intelligence. And here comes the more general claim: anything that varies between homes is already factored into that 5% of shared environment value. So if we took some children who were receiving a terrible such environment, say, 2 standard deviations below the mean (about 2th centile, i.e., worse than 98% of homes) and improved it to average, how big a difference would that make? The math is simple enough: the square root of 5% is (pulls out calculator) 0.22, times 2 because 2 standard deviations were improved, thus, 0.44 d. So the expected IQ increase would be about 7 IQ. This is the best one could hope for. Pessimism about interventions is thus based on the results of family studies. If the shared environment for something is 5%, then tinkering with such variables is not likely to be very productive because the best one can really hope for is small. Worse, realistically, we cannot do an intervention that would boost the home environment, insofar as intelligence is concerned, from the 2nd centile to the 50th. Parents are overwhelmingly responsible for their own children. We could do some improvement, say, give them a tutor for some hours a week, or make school time longer (assuming that school time is better than home time!). Worse, we don’t actually know what these 5% factors are. We only know there are some things that already vary between families and have a slight effect on intelligence. Thus, if we pick one random aspect of between-home variation and try to improve this in below average homes, this is not likely to produce any changes in the intelligence of the children, and it cannot produce big ones.

What about PKU, glasses etc.? The case of PKU — Phenylketonuria — is simple enough. It’s a genetic disorder that breaks the phenylalanine hydroxylase (PAH) gene so that phenylalanine builds up in the body which causes mental retardation. The genetic variation in PAH is part of the usual 80% heritability figure, which includes all genetic variation. However, PKU is extremely rare, so if we fixed this — gave every child with PKU perfect treatment for this — this would have essentially no impact on the variation in intelligence we see. Worse, we also know that our treatment for this is not perfect. Children who are on a PKU diet (i.e., not eating stuff with phenylalanine) are still some 10 IQ below average. This is because it’s extremely difficult to entirely avoid this amino acid. It is true that the PKU example shows that a heritability of 100% doesn’t prove that no intervention will work ever. 100% heritability does prove, however, that nothing that currently varies substantially between people can affect the outcome. In the case of high heritability, the only kind of intervention that can work well is one that introduces a new kind of environmental variation. In the case of PKU, this is a diet without phenylalanine, which didn’t exist before. In the case of eyeglasses, this is a new kind of wearable technology that counteracts incorrect configuration of the eyes for incoming light. Nothing in social policy is like this. Every social intervention tries to alter something that already varies between homes, whether it is number of books at home, reading to children, tutoring, diet etc. These social interventions we can already know from family studies cannot work well, and probably won’t work at all. Anyone taking a critical look at the reports of such randomized controlled trials will find that these generally don’t work. They can sometimes produce short-term changes in IQ scores, and probably not in intelligence. These will then fade away anyway, so there is no point to it. My point isn’t new at all. Sesardić makes the same point:

All this does not mean, of course, that heritable differences are written in stone, and that they are absolutely unmodifiable. It is always possible that something that was a purely genetic effect in a given set of environments will cease to be so when novel environments are introduced, and that it will then become responsive to these new environmental influences. But this possibility does not prove, as is frequently implied, that genetic effects are as easily modifiable as the environmental ones. The crucial structural asymmetry between the two may be described as fol- lows. Environmentally produced phenotypic variation can be leveled out by exposing all organisms to the same kind of environment, out of those already known. On the other hand, barring the policies that put environmental handicaps on those with a genetic advantage, genetically caused (heritable) phenotypic differences can be removed only if a new environmental difference is found outside the existing environmental range which will interact with genetic influence and which could be used to off- set it. Needless to say, in any particular case one simply cannot be certain that such an environment exists, or if it does that it is just about to be discovered.

I think this error of reasoning is another example of the deductivist fallacy. The deductivist fallacy is when one can prove mathematically or logically that one fact has no logically necessary connection to another fact (i.e., it is not self-contradictory to claim both). However, just about everything in science is actually about probabilistic reasoning, not necessity. Thus, while it is true that a high or perfect heritability for some phenotype does not logically rule out every imaginable intervention we might try, it does make them unlikely to work or at least be effective. That’s because almost all interventions, and probably all large-scale social interventions, involve equalizing some aspect of the environment that already varies between families, and thus is already included in the 5% shared environment component. If people who support interventions want to get serious, they need to start proposing interventions that change some aspect of the environment that doesn’t already substantially vary between families. Everything already varies a lot between families! What is there really to try that we haven’t tried already?