2. Mutational load

Mutational load, briefly put, is the effect that mutation has on the fitness of a population. It is one component of genetic load, which is the difference in fitness between the maximum fitness genotype in a population and the mean fitness of the population (Crow, Reference Crow and Kojima1970).Footnote ¹ Muller and Lynch’s arguments, however, are better understood as concerned with the expected number of deleterious mutations in an individual in a population, as opposed to, in the first instance, a difference in fitness.

Muller’s introduction to this concept is particularly clear. He imagines a delivery company that purchases two new trucks per year ( $n$ ). The trucks persist ( $p$ ) in the fleet until they wear out—3 years in the example—and are subsequently removed from the fleet. Eventually, an equilibrium is reached, and there is a constant number of trucks in the fleet ( $f$ ). In this case, that number is six. Thus, the number of trucks in the fleet at equilibrium may be found by the following equation:

$$f = np.$$

The same holds for deleterious mutations. Mutations enter the population at a rate ( $n$ ) proportional to the mutation rate, they persist ( $p$ ) in the population for a time inversely proportional to the strength of selection, and the expected number of mutations in the population per generation ( $f$ ) is determined by these values. Muller estimated, prior to the molecular revolution in genetics, that this number was eight in humans. Current estimates put the number at 12,000–13,000 single-nucleotide polymorphisms (SNPs), although, of course, their mean effect size is much smaller (Henn et al. Reference Henn, Botigué, Peischl, Dupanloup, Lipatov, Maples, Martin, Musharoff, Cann, Snyder, Excoffier, Kidd and Bustamante2016).

Given the previous formula, there are two ways to increase the mutational load of a population. The first is to increase the mutation rate. The second is to relax selection pressure.

Living, as he was, at the beginning of the atomic age, Muller was concerned with an increasing mutation rate due to atomic weapons and the proliferation of radioactive devices in medicine, industry, and the home. Crow (Reference Crow1997) speculates that had he been aware, Muller would also have been concerned by the spread of chemical mutagens. In this, he was undoubtedly right. An increase in radioactive and chemical mutagens is a cause for concern. It is likely to result in an increase in both somatic and germline mutations, and those mutations that affect fitness are, it is commonly assumed, overwhelmingly likely to be deleterious.Footnote ²

Lynch is also concerned about increases in the mutation rate, although his concern is primarily with increases in the germline mutation rate due to paternal age. It is well known that because the process of spermatogenesis involves the duplication of cells many times throughout a male’s lifetime, there is an increasing probability of mutations in sperm cells as males age. Thus, a trend toward older fathers increases the effective germline mutation rate.

The main target in both articles, however, is the relaxation of selection due to changes in the human environment. Lynch points out:

This is what makes humans “exceptional” as per the title of Lynch’s paper. We shape our environment and, in so doing, reduce selective pressures on mutations in the population.Footnote ³ This, given the previous formula, increases the number of deleterious mutations in the population.

3. The simple argument from load

The simplest version of the argument is as follows:

. This argument is sound, but more is needed to get the required biological conclusion:

The issue here is that an increased number of deleterious mutations does not necessarily correspond to a reduction in fitness. Fitness depends on selective environment as well as phenotype, but by hypothesis, the selective environment has changed, and thus we cannot immediately conclude that fitness will decline.

As a concrete example, consider type 1 diabetes. Type 1 diabetes typically develops in childhood, and in the ancestral environment, this would likely be a deadly affliction, resulting in a realized fitness of 0. But in our current environment, with early detection, blood sugar monitoring, insulin, insulin pumps, and so on, individuals with type 1 diabetes are increasingly likely to survive and reproduce. The expected number of offspring for an individual with type 1 diabetes now approaches that of an individual without type 1 diabetes (Sjöberg et al., Reference Sjöberg, Janne Pitkäniemi, Haapala and Tuomilehto2013). Fitness of the phenotype in the modern environment is not that of the individual in the ancestral environment, so it is false that the increased number of deleterious mutations causes a decrease in population fitness.

In fact, holding all else fixed, we can expect an increase in absolute fitness, or rate of growth, as a result of improved medical technologies. This is especially obvious given how common assisted reproductive technologies have become. Many of us have used or know someone who has used in vitro fertilization or other treatments for infertility. Many who would not otherwise have had children now can and do.

Ignoring the fitness effects of changes in the environment seems like an elementary mistake, but it is one Lynch seems to make in passages like the following:

This raises an obvious question: What is a deleterious allele when there is no selection? Lynch says that the effects are hidden, but the reference environment he uses is the ancestral one, not the current environment. Lynch needs to establish that the ancestral environment is, for some reason, a privileged reference environment and the one that ought to count when we consider the fitness of a population. “Hidden” fitness effects are simply not fitness effects in the current environment. As Pascoal et al. (Reference Pascoal, Shimadzu, Mashoodh and Kilner2023) observe in a study of parental care in the common burying beetle, Nicrophorus vespilloides, relaxed selection “results in a greater mutation load, for which it is also a phenotypic antidote.”

Roth and Wakeley (Reference Roth and Wakeley2016) point out that our current environment may be very different selectively from our past environment, with some traits that were advantageous in the ancestral environment being disadvantageous in the current environment, and vice versa. This is undoubtedly true for some traits, but Lynch replies that many others, such as psychological conditions and hereditary cancers, have “absolute effects in essentially all environments” (Lynch, Reference Lynch2016a). This may be correct—cancers seem unlikely to have positive fitness effects, although I am less confident about most psychological conditions—but it misses a larger point. That larger point is that environments with relaxed selection conditions like psychological disorders and hereditary cancers by hypothesis do not have the same fitness effects. This is not to deny that many traits are deleterious in both ancestral and relaxed environments—the claim is not that selection has been completely relaxed. The claim is that the selective effect of these conditions is less than it was in the ancestral environment, so they do not have the same fitness consequences.

Similarly, Lynch argues that even if selection is “soft,” deleterious mutations will still have “very real costs.” But it is unclear what these costs are. In hard selection, genotypes have the same fitness differences in different environments, such as a trait that reduces fertility in all environments, but in soft selection, population density makes a difference.

Wallace (Reference Wallace1991) used the example of hibernating bears to make this clear. If there are fewer dens to hibernate in than bears, being passive is a disadvantage—an aggressive bear gets a den over a passive bear. But if the number of dens is greater than the number of bears, being passive is no disadvantage at all—every bear gets a den regardless. Whether or not being passive is a disadvantage, and to what degree, depends on the selective environment, in this case, the population density.

Muller (Reference Muller1950) makes a similar mistake in his argument that an increase in mutational load would be detrimental for the human population, as Wallace (Reference Wallace1991) points out. Muller calculates that the decrease in human fertility due to load is approximately 0.2. Because of this, the 2.3 children per couple in Western societies (as of Reference Muller1950) represents the survivors of about 3 zygotes per couple. Doubling the load would then reduce the probability of survival per zygote from about 0.8 to 0.6, leading to only 1.8 children per couple, which is not sufficient for replacement and would lead to population decline and eventual extinction. What Wallace points out is that this ignores the fact that rates of reproduction in humans (especially in societies with access to birth control) are dependent on population density and are therefore soft.

If Lynch means that these conditions have constant costs across environments, then he is simply claiming that selection is hard. But that is not true for many of the conditions we might consider. Consider again the case of type 1 diabetes. Arguably, selection is soft on this trait because it depends on the availability of medical care. If care is expensive and rare, then there will be more intense selection pressure on the trait. But if care is cheap and common, there will be less selection pressure, if any, on the trait. Obviously, the cost and availability of care vary by condition, so Lynch is right that selection is hard for some traits. But this only shows that the selection against these traits is not as relaxed as it might be. If selection is stronger than we thought, then the expected increase in mutational load is also lower. One is left wondering what Lynch means when he says that there are “very real costs.” If they are fitness costs, there is less load; if the costs are something else, what are they?

4. The argument from declining baseline human performance

In order to make the argument that relaxed selection poses a problem for human populations, there needs to be a bridge from the claim that there will be an increase in the number of deleterious mutations to the claim that something bad for the population will happen. In a second version of the argument, rather than a decrease in fitness, an increase in the number of mutations is expected to result in a reduction in “baseline performance.” Lynch makes this claim, saying:

This claim is based on mouse studies by Uchimura et al. (Reference Uchimura, Higuchi, Minakuchi, Ohno, Toyoda, Fujiyama, Miura, Wakana, Nishino and Yagi2015) in which a mutator strain of mice was seen to decrease in offspring number and body weight and increase in “obvious phenotypic abnormalities.” Lynch notes that mice and humans have similar genetic and genomic architectures, and thus he holds that the inference from mouse to human is valid. This is too quick, however. Even if the inference from mouse to human holds, the study by Uchimura et al. (Reference Uchimura, Higuchi, Minakuchi, Ohno, Toyoda, Fujiyama, Miura, Wakana, Nishino and Yagi2015) is not a study of relaxed selection; rather, it is a study of increased mutation rate. Mutator strains of mice are strains with impaired DNA-repair mechanisms and thus have a higher mutation rate than wild mice. As we’ve seen, relaxing selection is likely to increase offspring number, not decrease it, because previously less fit phenotypes now have higher numbers of offspring.

If there is a signal to be found here, it is that we will observe non-fitness differences, such as changes in body weight and “obvious phenotypic abnormalities.” But in order to assume that these will affect baseline human performance, we would have to know what human performance metrics we’re looking at and whether these are or have until recently been under selection.

So what performance metrics does Lynch consider? The primary one that Lynch considers is intelligence. He mentions studies from Crabtree (Reference Crabtree2013) and Woodley of Menie (Reference Michael2015) that suggest there has been a decline in intelligence in the United States and the United Kingdom and that, although controversial, these claims give us reason to think that there is a problem.Footnote ⁴ Lynch is right that these studies are controversial; indeed, they are extremely problematic for several reasons. Woodley of Menie (Reference Michael2015), for example, uses socioeconomic status as a proxy for intelligence, then points to a historical association between socioeconomic status and realized fitness, or actual offspring number, to indicate that intelligence has a fitness effect. Unpacking all that is wrong with this would take some time, but the real problem for Lynch is elsewhere.

The real problem is that these articles assume the conclusion that Lynch wishes to draw. Both of the cited articles take it as given that mutational load is affecting intelligence and attempt to estimate the amount of decline that has occurred. But this is exactly Lynch’s argument: Selective pressure has been reduced, and therefore there has been (or will soon be) a decline in intelligence. Far from providing independent support for the claim that intelligence has decreased or will decrease, it makes the very same argument that Lynch is making.

There’s also the question of what the 1% decrease is with respect to. The way the worry is phrased, it might sound like we’ll see a 1% decline in hundred-meter dash times, maximum bench-press strength, or measured IQ. This isn’t the case. The claim is that fitness will decline by 1% per generation due to the spread of deleterious mutations. That is, it’s a claim about a decline in fitness from the improved baseline after intervention and not about declines in phenotypic measures of performance.

Furthermore, the mapping from fitness measures to phenotypic measures isn’t at all obvious. What decline in speed, strength, or IQ would equate to a 1% decline in fitness, and in what selective environment?Footnote ⁵ What aspects of strength or intelligence are under selection in the first place? Much more would need to be said before we can conclude there is a significant worry here.

Now consider Lynch’s claim that for traits that were under selection, we can expect to see an increase in “obvious phenotypic abnormalities.” Here again, more needs to be said. Why are these “abnormalities” problematic? Eyeglasses—and more recently, laser surgery—could well be reducing selective pressure on visual acuity.Footnote ⁶ If that’s right, we should expect an increase in the proportion of people needing glasses. But as James Crow pointed out with respect to this question, “Who worries about having to wear spectacles?” (Crow, Reference Crow2000). Crow’s point applies more generally. Advances such as insulin pumps and continuous blood glucose meters mean that where such care is widely and cheaply available, outcomes for type 1 diabetes continue to improve (Beck et al., Reference Beck, Bergenstal, Laffel and Pickup2019). Soon we may be able to say, “Who worries about having to check their blood sugar?”

In order to make this point stick, Lynch would have to show that some important good would be threatened if these conditions were more common. But doing this would move away from the apparently value-neutral—“presenting the facts as they are”— argument Lynch seems to want and into explicitly ethical territory. By framing the question in terms of “baseline human performance,” I take it that Lynch tries to stay in the biological realm while also making a covert value judgment that these performance metrics are or ought to be worrying.

5. The argument from the long term

This brings us to the third argument that mutational load is a problem in need of a (eugenic) solution. Lynch argues that solutions like eyeglasses and insulin only exacerbate the problem (Lynch, Reference Lynch2016b). They further relax selection, leading to a greater need for medical intervention. But there’s no guarantee that medical technology will keep ahead of the curve.

This argument differs from the previous ones in an important respect. Here, the argument isn’t that fitness declines with respect to the ancestral environment. The argument is that fitness will decline with respect to the improved baseline after the introduction of improved sanitation, medicine, and so on. After a sharp jump up, there will be a slow decline back to mutation-selection equilibrium—the point at which the introduction of deleterious mutations equals the rate at which they are removed by selection.

Muller makes the same claim in his 1950 article. He points out that relaxing selection makes no difference in terms of “genetic deaths.” At mutation-selection equilibrium, population fitness depends only on the mutation rate. This is because the effect of a deleterious mutation on the population is the same no matter how weakly or strongly deleterious the mutation is. The only difference between a weakly deleterious mutation and a strongly deleterious one is that a weakly deleterious mutation will end up affecting many more individuals than a strongly deleterious mutation before it is removed from the population. Eventually, a new equilibrium will be reached, and natural selection will be back in full force. Our efforts to cheat natural selection through technology will have been in vain. That’s why Muller concluded that the only way to avoid a return to full-force natural selection was to take the more “humane” route of artificial selection—in other words, eugenics.

Lynch builds on this, arguing that this is not merely about the introduction of new mutations into the population. Even if no new mutations were introduced, there would be a reduction in fitness due to preexisting deleterious mutations drifting to higher frequencies (Lynch, Reference Lynch2016a).

While this is all correct, it is a long-term problem and does not show that medical intervention was futile. Lynch estimates that if the selection coefficient of a mutation is reduced from 0.01 in the ancestral environment—which he estimates to be the usual size of a deleterious mutation—to 0.001 in the new environment, the halfway point will be reached in about 700 generations if no further advances in medical technology and availability are made. Until equilibrium is reached, there will be a reduction in “genetic deaths” from natural selection. Is the fact that we’ll eventually reach equilibrium a reason that the reduction in selective force until then was “futile”? Postponing or pausing a problem is not a bad thing.

We’re all familiar with Keynes’s quip that “in the long run we’re all dead.” His point was in reference to the quantity theory of money. According to that theory, increasing the amount of money in an economy makes no difference because eventually prices will simply rise to match. Real prices will not have changed at all; therefore, injecting money into an economy is futile. But as Keynes points out, even if this is true in the long run, we may care very much about the short-term changes that occur. The same can be said for the short-term changes that occur through medical and other technological interventions.