Contra Seb Jensen on Mutational Load
I think his article is overall weakly in favor of my theory of mutational load
Occasionally people talk about mutational load, and they mention me as a proponent as the theory, and then if they aren’t themselves for the theory they cite something they think stands against the theory. Recently, Emil Kirkegaard and Werner Zagrebi referenced a post by Seb Jensen. A lot of the stuff in his post comes from other people’s criticism articles, which I’ve addressed elsewhere, but not all of that made it to Substack. So I figured it might be time for another article on mutational load theory, since Seb’s article is so relatively popular.
What I believe
There are a lot of misunderstandings of mutational load theory that reccurently pop up. These include:
Mutational load is a meltdown theory that says society will soon end
Mutational load is caused by older fathers and men shouldn’t breed after ~25
Mutation rates are higher post industrialization
Mutational load is the unique cause of ${dysgenicTrend}
I don’t believe any of these things. Rather, I believe mutational pressure, as contrasted with selection pressure, gene flow, and genetic drift is an important force in human evolution. It means that, even with large population sizes, if there are no selection pressures (meaning everyone breeds the same amount, or they breed independent of their traits), important traits will still deoptimize to a significant extent each generation. This doesn’t cause meltdown, just dysgenics. It’s somewhat slow if your reference point is 0 — mutational load maybe causes the loss of 1 or 2 IQ points at the mean per generation. Low selection pressure is the much more important factor. Mean paternal age doesn’t matter if at every paternal age, smarter fathers are breeding more. The same is true for mutation rates. These don’t need to go up with industrialization; rather, selection pressure came down. Likewise, dysgenic trends in large populations, defined as experiencing no gene flow, can be understood as the sum of mutational and selective pressures.
An epistemic warning
I spun this section out into this article. Shortly said, I’d like you to consider the mathematical complexity of mutational load theory and how talking about it verbally reduces the ability of the expert to make a solid case. The mathematical case may be a lot more solid than the verbal case — but only the verbal case is accessible to the layman. So the theory underperforms in the marketplace of ideas, without even considering intrinsic political bias against it. This is the case with mutational load theory — it is double cursed as well, as there is almost certainly political bias against it. Probably even among the right.
Seb’s signpost
Now let’s get started. We’ll follow Seb’s signpost:
Here is the evidence that mutational load is not that important:
The French are not uniquely mutationally loaded in comparison to third world populations.
The (admittedly opaque) simulation studies suggest that low purifying selection is not an issue because the mutation rate is much less important than the balance between beneficial and deleterious mutations.
The declines from the negative Flynn effects are way too large to be due to mutational load, and are likely due to non-g variance or immigration.
Controlling for birth order effects attenuates the relationships between parental age and fitness indicators such as intelligence.
Out of the five parental age studies on IQ that exist, two have statistically insignificant negative effects after controlling for birth order.
We have not lost our heads or our money after over a hundred years of industrialization.
There are several plausible purifying selection mechanisms that still exist: abortion, miscarriages, sperm selection, natural selection, and sexual selection.
The French are not uniquely mutationally loaded in comparison to third world populations
There are issues with both the validity of this argument and the truth of the premises. Let there be n genetically disjoint traits. Mutationally load theory claims there are 2 or 3 disjoint traits where the effect on the phenotype is greater than the negative effect on fitness in modernity. These are leftism, IQ, and mental illness. Mutations accumulate in these trait areas. Other populations in the world seem to be under dysgenic selection for IQ, so perhaps they also accumulate more mutations in IQ, the less IQ matters in their selection environments. They may also accumulate more mutations in some other behavioral trait like GFP, which could be under stronger selection in Western societies. The overall amoun of mutations can balance out, even though the qualitative texture of what is mutating differs between populations. Thus, strictly speaking, mutational load theory as advocated by me makes no predictions whatsoever about “number of derived alleles relative to Yoruba”, which is the specific metric used by Seb’s citation. Due to the invalidity of this argument, we can throw it out as completely uninformative a priori.
Considering again my epistemic warning, I’ve probably lost some people with this logical argumentation. So in addition I will say that the factualness of this claim is also debateable. Another paper found conflicting results.
Here TSI (Italians) have more mutated variants than Yoruba. Japanese have even more. Seb’s citation says on this:
The contradictory conclusions about the impact of recent demographic events on the mutation load in humans can be traced back to the use of different summary statistics and to how they are affected by factors other than load. We contend that of the summaries used thus far, the number of derived alleles per individual is both more sensitive and specific to load.
This contention is arbitrary.
Of course, future advances such as improvements in our classification of variants, may reveal differences in load that evade us at present. However, current findings suggest that differences among extant human populations are likely to be small.
I will end off with a right-wing political heuristic. This is also a race denial paper. The whole point of inventing the new summary statistic to contradict previous research was to minimize genetic race differences. It’s fair to say that the paper’s authors were likely motivated by race-leftism in writing their paper. Without knowing much else, this may motivate you to look at this paper with more skepticism, if you understand how this sort of motivation is often the source of a lot of bad and obscuring science.
The (admittedly opaque) simulation studies suggest that low purifying selection is not an issue because the mutation rate is much less important than the balance between beneficial and deleterious mutations.
First, as far as I can see it’s one study, not a number of “studies.” Second, we are concerned with the effect of mutational on phenotype, not on fitness. This simulation claims that a model in which one mutation confers a large beneficial increase, making up for 10 deleterious mutations of small fitness decrease.
While this simulation in no way demonstrates that this is the case in reality, if you do a similar model over a similar time period as that from 1850 to now, you get that the median decreases, the average goes up a bit, and variance increases.
So you still get phenotypic decline at the median, while the species stays the same. Importantly, we don’t see the deformation of the bell curve or the variance explosion that this model implies — these facts may be enough for the model to be “put to rest”, in the sassy and overconfident words of the authors of the articles’ abstract.
I think it’s clearly best if this study is completely dismissed for being unempirical, having unrealistic assumptions, not even implying the lack of phenotypic decline at the median due to industrial mutational pressures, and being not about qualitative phenotype but rather fitness.
The declines from the negative Flynn effects are way too large to be due to mutational load, and are likely due to non-g variance or immigration.
Seb is referencing the chart above. To begin, IQ changes having an environmental component does not rule out mutational load theory at all. Mutational load theory is a genetic theory. I employ it to explain the rise in leftism over the 20th century. IQ rose over the 20th century, and this was in spite of dysgenic selection and mutational load. Therefore, something in the environment must have effected IQ scores tremendously. As such, the negative Flynn effect is of little concern; what we are really interested in is the robust causal paternal age correlation with IQ. That, or a molecular estimation of the mutational effect.
We have a molecular estimate I discuss here.
I discuss in this article that this might be an underestimate as they only look at the exome and it comes in to about half the size of our best (admittedly somewhat bad) paternal age effect estimates. So, let’s take the left end of the bell curve and cut it in half to get an upper bound on the mutational effect per decade — 1.5 IQ points. This works in tandem with dysgenics, which Seb’s own meta-analysis estimates can decrease IQ by .5 points per year. So we actually explain more than 70% of 5/8 declines with dysgenics and mutational load.
The rest must be explained with something else. Seb mentions immigration, which is definitely decreasing national IQ. It is a serious flaw with regards to studying change in a population if recent immigrants are included in the data. They are by definition not members of that population. They are members of another population that have moved in. Only the mixed race offspring of intermarrying immigrants should be considered to be members of the population in question. Only upon intermarriage does gene flow really occur. And intermarriage rates are quite low; they may explain some of the decline but what is likely a larger effect is simply including 1st and 2nd generation immigrants in samples.
Seb’s point here provides no evidence against mutational load theory. Rather, it provides evidence against the claim that mutational load is solely responsible for declines in IQ. I personally have never made such a claim — I only make such a claim regarding leftism in whites. Although I would not be surprised if mutation and dysgenics together explain most of the decline (>70%) within whites. Given that immigrants are present in Seb’s data, this hypothesis cannot be ruled out.
Controlling for birth order effects attenuates the relationships between parental age and fitness indicators such as intelligence.
I believe this section more or less all comes from Cremieux, who I already replied to on Substack. The TLDR is that this data is consistent with a negative paternal age effect and environmental effects coexisting. As such, Cremieux convinced me there is an environmental component to the IQ birth order effect, and that it must be controlled for in regressions on paternal age. When it is controlled for, the environmental effects of birth order are filtered out, so our estimate of the mutational pressure is not confounded by these effects. So what really matters here is what the paternal age effect is when controlled for birth order (i.e., what happens to first born IQ when dad goes from 20 years old to 40 years old).
Out of the five parental age studies on IQ that exist, two have statistically insignificant negative effects after controlling for birth order.
These studies must use family fixed effects with controls for birth order, or they are confounded, since older fathers tend to be higher IQ genetically. Seb says he cites 5, but in the article I find only three. All of them use family fixed effects or control for family PGS but one doesn’t control for birth order. That one still found a confidence interval consistent with the hypothesized effect (in between -.5 and -.1 SDs). It was, however, statistically insignificant, meaning the study was underpowered and thus inconclusive.
The next one is the same story. At this point we can’t rule out or verify mutatonal load on IQ.
The final citation is also underpowered but the hypothesized trend fits within the fixed effects confidence intervals, the fact that they did not measure IQ but rather either used EA or a binarization of IQ aside.
Instructive is that this same study found a robust pattern of negative paternal age effects on mental illness, controlling for birth order and using family fixed effects.
Combined with the molecular study I cited earlier, this makes it very hard for me to believe there is no significant paternal age effect on IQ and leftism.
So overall I would say this section should give weak positive evidence for mutational load theory.
We have not lost our heads or our money after over a hundred years of industrialization.
Mutational load theory is not a catastrophe theory. Still, there is Woodley’s data on reaction time (Seb cites this), the decline of genius, and the g-hollowness of the Flynn effect. People are undeniably different than those who lived in the 19th century, in the direction of being less virtuous and more intellectually insipid. Simply put, mutational load theory does not say that society should have collapsed by now. Rather, it attempts to explain exactly the decline that we have observed, not more than that.
There are several plausible purifying selection mechanisms that still exist: abortion, miscarriages, sperm selection, natural selection, and sexual selection.
Natural selection is dysgenic for IQ and so won’t purify a mutation that only decreases IQ. Natural selection is generally weak. Sexual selection is not mathematically distinct from natural selection in this case as natural selection is defined as the correlation between fertility and traits. Abortion likewise; if people with certain traits choose to have more abortions, this is reflected in the data by them simply having lower fertility.
Miscarriages and sperm selection are more properly considered as purifying selection. I have looked into the miscarriage rate before and devised an equation in here to correct for it as well as infant mortality. It did not end up really mattering, because these are a combination of rare and not correlated with paternal age. Besides this, it turns out that if you look at offspring, they actually do get the expected number of de novo mutations given their father’s age. In other words, men accumulate about 2 mutations per year in their sperm, and these actually do make it through to the offspring. So they are not being purified much prior to natural selection.
Seb specifically says on this:
The current method of estimating the effect mutagenic dysgenic effects on population level phenotypes is estimating the causal effect of parental age (e.g. decrease of 2 units per year of paternal age) and then multiplying that by 100 to obtain the expected change per century (200). This model works, and only if, there is zero purifying selection for deleterious mutations. It does not if there is selection against these mutations due to sperm selection, miscarriages, infant mortality, natural selection, sexual selection, or genetic drift. While the purifying selection against mutation is probably falling in infants, it is unclear how much purifying selection has declined due to lowering rates of infant mortality.
This isn’t my model!
Also note that claiming purifying selection is purging mutations contradicts the basis of the simulation from earlier. These two points cannot coexist in the same argument as they contradict each other.
Conclusion
Let’s review every point.
The French are not uniquely mutationally loaded in comparison to third world populations.
This came to nothing as the literature was both irrelevant and inconsistent. Effect: 0
The (admittedly opaque) simulation studies suggest that low purifying selection is not an issue because the mutation rate is much less important than the balance between beneficial and deleterious mutations.
This also came to nothing as the literature is unempirical, potentially not consistent with real world assumptions and observations, and fails to contradict mutational load theory. Effect: 0
The declines from the negative Flynn effects are way too large to be due to mutational load, and are likely due to non-g variance or immigration.
This contradicts a different theory that mutational load explains all intelligence declines. But I would say the effect on the mutational load theory I hold is 0.
Controlling for birth order effects attenuates the relationships between parental age and fitness indicators such as intelligence.
This contradicts a different theory that mutational load explains all intelligence birth order effects. But I would say the effect on the mutational load theory I hold is 0.
Out of the five parental age studies on IQ that exist, two have statistically insignificant negative effects after controlling for birth order.
I would say the effect here ranges from 0 to weakly positive for mutational load theory, especially because of the effects on mental illness in one of the citations. Effect: 0 to weakly +
We have not lost our heads or our money after over a hundred years of industrialization.
This contradicts a different theory that mutational load predicts fast civilization collapse. But I would say the effect on the mutational load theory I hold is 0.
There are several plausible purifying selection mechanisms that still exist: abortion, miscarriages, sperm selection, natural selection, and sexual selection.
This addresses a model I don’t use, but its effect on my model for estimating the effect should be 0, as I already account for purifying selection as well as natural selection when giving overall evolutionary pressure. Effect: 0.
Our aggregate effect is 0 to weakly postive for mutational load theory. However, I think the article successfully rebuts a different theory that mutational load explains all IQ decline, all birth order effects, and will cause fast civilization collapse.
I don’t think the theory that mutational load causes a decline in 1 to 2 points of IQ per generation can be rebutted presently, nor can the twin theory that there is an effect of the same size on leftism. I think that overall, if you consider all of the evidence, the effect should be moderately positive. But this does not mean the theory is proven. I give it about a 95% chance of being true. If the theory is not true, my next best theory for the rise of leftism is the Flynn effect, which was so strong it covered up the genetic decline of IQ. But that’s a post for another time.
Inclined to think about this with animals somewhat.
Ive seen over a 300k pigs get born over 5 years and noticed results and mutations increase after 6th births round.
Most are physically noticeable but theres behavioural patterns including antisocial behaviours as fighting and self isolation which i rarely see with prime productive birth round (3rd) or first round (usually low count but high quality pop)
In general would it not be easier to prove these things with animals.
Might not be capable of gathering IQ stats lol, but thats not what this is about at its core.
Ofcourse that does leave up for more ideological and tactical denial, hur man isnt animal types, or tactical proving effect on animal doesn't on man.
Some of these arguments against mutational load are terrible.
Argument 2: Mutational load is not about the total number of mutations. It's the number of bad mutations (and their effects) that matter. Purifying selection doesn't even necessarily decrease the total number of mutations–the number of mutations is almost exactly determined by paternal age, which could very well be (mostly) independent of the strength of purifying selection.
Argument 6: The qualitative effects of mutational load depend on the coefficients of selection. It's not "mutational load is true therefore society will collapse". Rather, it could be "mutational load theory predicts -10IQ per decade, hence society will collapse". But obviously mutational load is not predicting -10IQ per decade.