Does fertility start to decrease beyond marriages with fourth cousins?

My reply to Razib Khan

Is the past an action-packed story of massive turnover, admixture, and periodic mass migrations? No, not mostly. (Wikicommons: Guillaume Lardier)

Razib Khan has written an interesting reply to my piece on outbreeding, particularly my contention that fertility starts to decrease beyond marriages with fourth cousins (Frost, 2024; Khan, 2024). This decrease, I argued, is “the canary in the coal mine.” It’s the first sign of something going wrong when outbreeding brings together different gene systems.

In my piece, I reviewed the evidence on outbreeding and decreased fertility:

• Helgason et al. (2008) used a genealogical database to examine data on 160,811 Icelandic couples born between 1800 and 1964. This study provided evidence of both inbreeding depression and outbreeding depression, with fertility peaking at marriages between third or fourth cousins. The children of such marriages also had the same peak fertility. The authors divided their data into 25-year intervals to investigate the possible effects of social changes over time. Those short intervals still showed an inverted V-shaped association between fertility and parental relatedness.

• Laboriau and Amorim (2008) used the Danish Central Personal Register to examine data on 22,298 Danish women born in 1954 and still living in that country in 1969. This study likewise provided evidence of both inbreeding depression and outbreeding depression, with fertility peaking at marriages between partners whose home parishes were 75 km apart. The relationship between fertility and distance was not explained by education, family income, urbanicity, or mother's age at first birth.

• Joffe (2010) argued that an increase in outbreeding explains a century-long decline of sperm counts and a corresponding increase in testicular cancer. This explanation seems to provide a better "fit" to the geographic and temporal distribution of the sperm count decline than the more popular explanation of increased exposure to estrogenic compounds.

• Davenport and Steggerda (1928) administered psychological tests in Jamaica to 300 adults and some 1,200 children of Black, White, and biracial origins. The biracial participants had results midway between those of the Black and White participants—with the exception of certain psychological tests where they had to compare or imagine objects in two or three dimensions. On such tests, their failure rates exceeded those of the other two groups.

Keep in mind that decreased fertility is just that—your chances of having a baby are reduced by a higher risk of embryonic or fetal miscarriage. In most cases, persistence will pay off. Nonetheless, those embryos and fetuses are miscarrying for a reason, and the reason seems to be an increase in genetic incompatibilities with decreasing parental relatedness.

An extreme example is provided by crossings between modern humans and archaic humans, like the Neandertals. At some point, Neandertal genes entered our gene pool, either directly 50,000 to 60,000 years ago when modern humans left Africa or indirectly as early as 120,000 years ago via an intermediate population in the Middle East (the Skhul-Qafzeh hominins). Whatever the scenario, much of that archaic ancestry has been selected out of our gene pool, particularly genes associated with fertility:

One line of evidence for reduced fertility in male hybrids is that the proportion of archaic ancestry in modern humans is significantly reduced on chromosome X compared to the autosomes. This is suggestive of reduced male fertility as loci contributing to this phenotype are concentrated on chromosome X in hybrids of other species. We confirm an extreme reduction of Neanderthal ancestry on chromosome X (16%–34% of the autosomes depending on the population) and find a quantitatively similar reduction of Denisovan ancestry (21% of the autosomes in Oceanians).

The second line of evidence in support of the hypothesis of reduced fertility in hybrids is that there is a reduction of archaic ancestry in genes that are disproportionately expressed in testes, a known characteristic of male hybrid fertility. (Sankaraman et al., 2016)

This reduction of archaic ancestry was not confined to genes associated with fertility. Petr et al. (2019) conclude that “selection against [Neandertal] introgression was strongest in regulatory and conserved noncoding regions compared with protein-coding sequence (CDS).” It seems that the incompatibilities were not in the genes that manufacture the proteins of human tissue. Rather, they were at a higher level—in genes that regulate growth and development.

Eventually, those incompatibilities were purged from the gene pool through successive generations of purifying selection. How many generations? Petr et al. (2019) estimate ten or so: “Similar to Harris and Nielsen, we observed abrupt removal of Neandertal alleles from the modern human population during the first ~10 generations after introgression, followed by quick stabilization of Neandertal ancestry levels.” That estimate, however, is based on the assumption of a single introgression some 55,000 years ago. Many academics argue for multiple introgressions that go farther back in time.

But let’s suppose that estimate is right. Should we feel reassured knowing that any harmful effects of outbreeding will be purged from the gene pool in ten generations? Keep in mind how much stronger natural selection was 55,000 years ago. Today, with infant mortality largely gone and fetal mortality reduced, the gene pool would take much longer to purge itself. 

At this point, you might ask: “But wouldn’t there be a lot less to purge in crossings between modern human populations?” Yes, of course. But “a lot fewer” doesn’t mean “almost none.”  We could try to estimate the load of genetic incompatibilities by measuring the purifying selection in introgressions from one modern human population into another. Has anyone actually done this?

According to Razib Khan, someone has:

In 2014 I specifically asked a postdoctoral fellow in David Reich’s laboratory if they had looked to see if the pattern with Neanderthal introgression could be seen more modestly in the African forager genomes that they had early access to through the Simons Foundation, and was told they had detected no such signature. (Khan, 2023)

This “study” has three shortcomings:

1. It’s inside information. As such, it’s difficult to evaluate, all the more so because the source is unnamed. This is why research findings are normally published in academic journals, where the author has to provide the data, explain the methodology, spell out the assumptions, and acknowledge the shortcomings. After publication, the author should be available for any questions.

   The problem here is not that an unnamed source may lie (although that does happen). The problem is that an unnamed source may speak in good faith and still be wrong. It’s often only when you put pen to paper that you begin to notice the faults in your reasoning. More faults may become evident during peer review and then after publication. Without that process of internal and external verification, an opinion—even if given in good faith—may be tainted by groupthink and unexamined assumptions.

2. The methodology is imprecise. The Icelandic and Danish studies directly measured the effects of outbreeding on fertility. This kind of methodology is generally superior to one where the effects are estimated by using a model where each calculation is plagued by uncertain assumptions and margins of error.

   It would be easier to measure the purifying selection after the Neandertal introgression into the human gene pool than, say, after the pre-Bantu and Bantu introgressions into the Khoisan gene pool. The Neandertals died out some 40,000 years ago, so there is no need to control for subsequent admixture. Such controls are needed, however, in the case of the Khoisans, who have been in contact with other populations up to the present. Yes, there are ways to remove recent admixture from the calculations, but they are subject to error.

3. The genome of the introgressing population is known only approximately. The Neandertal genome has been fully sequenced thanks to DNA from 50,000 years ago—which is close enough to the presumed introgression at 55,000 BP. In contrast, we have not yet reconstructed the genomes of the various Bantu and pre-Bantu groups that introgressed into the Khoisan gene pool over the past 150,000 years. Some of those groups have died out, and not enough of their DNA has been retrieved to reconstruct their genomes. The Bantu are still around, but they have undergone significant genetic change in the meantime. Keep in mind that Africans are not “living fossils.” They have pursued their own trajectories of evolution, just like humans elsewhere (Frost, 2023).

In sum, it is difficult to measure the purifying selection that follows an introgression from one modern human population into another. There is too much noise in the data. We can measure the purifying selection after the Neandertal introgression because the impact on human fertility was much more dramatic, and thus easier to make out against the noise. Also, we have fully sequenced the Neandertal genome. This is not the case with the genomes of most human populations that have existed in the past. The main obstacles here are lack of funding (and interest) and the difficulty of retrieving ancient DNA in tropical environments, where DNA degrades much faster.

Razib Khan’s take on the Icelandic study

In my opinion, the best avenue for future research would be to replicate the Icelandic and Danish findings by using another genealogical database, like BALZAC in French Canada.

Razib Khan does mention the Icelandic study but rejects its findings because the genetic differences between degrees of consanguinity are “minimal.” He again refers to an unnamed source:

I asked a friend who has worked more recently with Icelandic genomic data about that conclusion, and his intuition was the same as mine: the genetic differences between this kinship group and those further out are minimal, so it is likely not a biological dynamic. (Khan, 2023)

How “minimal” are these genetic differences? How many more genes become different, on average, with each successive degree of consanguinity? Unfortunately, we don’t know, since that chart has never been plotted.

Razib Khan may be referring to the finding that, on average, any two individuals are 99.9% genetically identical—as measured by nucleotide sequences. So what’s left to measure? Those same sequences, however, also show that humans and chimpanzees are 98 to 99% genetically identical. Is that difference minimal?

Keep in mind that a gene is an assemblage of nucleotide sequences. So each and every human gene could be 2% different from its chimp counterpart, and a 2% difference can significantly alter how each and every gene functions. Furthermore, the human-chimp difference rises to 4-5% if we factor in other differences between the two genomes, particularly in deletions and duplications (Suntsova & Buzdin, 2020). Finally, if we disregard nucleotide sequences entirely and look at the products of genes, i.e., proteins, we see that 80% of them differ between humans and chimps:

The chimpanzee is our closest living relative. The morphological differences between the two species are so large that there is no problem in distinguishing between them. However, the nucleotide difference between the two species is surprisingly small. The early genome comparison by DNA hybridization techniques suggested a nucleotide difference of 1–2%. Recently, direct nucleotide sequencing confirmed this estimate. These findings generated the common belief that the human is extremely close to the chimpanzee at the genetic level. However, if one looks at proteins, which are mainly responsible for phenotypic differences, the picture is quite different, and about 80% of proteins are different between the two species. (Glazko et al., 2005)

We are only starting to quantify the number of genes that differ between different human groups, and no one has yet quantified the number that differ at each degree of consanguinity. Those numbers may be much larger than what we might expect. Even identical twins differ on average from each other at 5.2 genes (Jonsson et al., 2021).

Broader questions

Finally, let’s go beyond the methodological issues and address broader, more epistemological ones: How much do we really know, and what remains to be known? Should we, in fact, pursue this line of enquiry? Is it legitimate? And why should kinship matter anyway?

How much do we really know? Even if one accepts that fertility starts to decrease beyond marriages between fourth cousins, the decrease might be nothing to worry about. Most couples wish to have only two or three children, and that target seems attainable for the range of consanguinity shown by Icelandic and Danish couples. But consanguinity is usually a lot lower for spouses in other Western countries, and their fertility might be more compromised. This is, in fact, one of the reasons offered for the decline of sperm counts throughout the West. Unfortunately, we don’t know. Nor do we know whether the higher risk of developmental failure affects not only the embryonic and fetal stages but also later stages after birth. All of these risks remain uncertain in both their magnitude and their nature.

We need more data on the inverted V-shaped relationship between consanguinity and fertility. Does fertility always peak at marriages with third or fourth cousins? And does it decline at a steady rate further out?  Or does the decline taper off? This relationship could be explored in other genealogical databases, like those from French Canada, Finland, and Estonia. In this, we should clearly distinguish between outbreeding depression and inbreeding depression, since the genetic problems are qualitatively different.

Is this line of enquiry legitimate? Whatever one’s politics, there is a widespread feeling that personal choices in sexual relationships are just that: personal. They are not to be questioned by others, even on an abstract level. This view is widely shared by libertarians on the right and by autonomists on the left. It is not incompatible, however, with the view that personal freedom is served not only by a wider range of choices but also by more and better information to make them. Ultimately, freedom is best served by informed choices.

Why should kinship matter? We cannot understand outbreeding without understanding kinship—a concept that has become alien to many of us. A term like “degree of consanguinity” means nothing unless we know who our ancestors were and who our cousins are. But how many of us do?

At the University of Washington, anthropologist Pierre van den Berghe would each year ask his new students to name all four of their grandparents. Most couldn’t. His finding has been confirmed by another American study, which found that only 47% of the respondents could correctly name all four of their grandparents (Melore, 2022). I myself would have trouble naming all four. It was only when I came to Quebec that I encountered tons of people who could name not only their grandparents but also their cousins. They valued kinship as something that concerned them personally, and they were just ordinary people. 

Without that understanding of kinship, issues relating to consanguinity will be understood only in a vague, abstract sense. The same goes for understanding of history and prehistory or, conversely, the post-kinship world of today—where the vast majority of social interactions are transient, impersonal, and with non-kin. We may thus fall prey to sensationalized accounts that tell us more about the present zeitgeist than about past reality. This is what I call “The Indiana Jones School of Anthropology,” and it comes up in Razib Khan’s article: “The ancient DNA makes it clear that the human past was defined by massive turnover, admixture, and periodic mass migrations of males.”

Well, no. The past was mostly rather dull. During the entire span of human existence, our ancestors largely lived among kith and kin who looked and behaved pretty much like they did. Life was indeed defined by “an equilibrium state of organically developing kinship networks.” Yes, those networks were disrupted at various times and places, but those disruptions were the exception, not the rule, and they were seared into human memory precisely because they were exceptional.

References

Davenport, C.B. & Steggerda, M. (1928). Race Crossing in Jamaica. Washington: Carnegie Institution, Publication no. 395. http://www.velesova-sloboda.info/archiv/pdf/davenport-race-crossing-in-jamaica.pdf    

Frost, P. (2023). The Ghosts of Africa. Africa had its own Neanderthals. Aporia Magazine, September 7.

Frost, P. (2024). The Goldilocks zone between inbreeding and outbreeding. Peter Frost’s Newsletter, August 20.

Glazko, G., Veeramachaneni, V., Nei, M., & Makalowski, W. (2005). Eighty percent of proteins are different between humans and chimpanzees. Gene, 346, 215-219. https://doi.org/10.1016/j.gene.2004.11.003  

Helgason, A., Pálsson, S., Guðbjartsson, D.F., Kristjánsson, þ., & Stefánsson, K. (2008). An association between the kinship and fertility of human couples. Science, 319(5864), 813-816. https://doi.org/10.1126/science.1150232

Joffe, M. (2010). What has happened to human fertility? Human Reproduction, 25(2), 295-307. https://doi.org/10.1093/humrep/dep390   

Jonsson, H., Magnusdottir, E., Eggertsson, H. P., Stefansson, O. A., Arnadottir, G. A., Eiriksson, O., ... & Stefansson, K. (2021). Differences between germline genomes of monozygotic twins. Nature Genetics, 53(1), 27-34. https://doi.org/10.1038/s41588-020-00755-1

Khan, R. (2024). The Genetic Time Machine: Neanderthals, genomics and hybridization priors. Aporia Magazine, January 16.

Labouriau, R., & Amorim, A. (2008). Comment on "An Association Between the Kinship and Fertility of Human Couples." Science, 322(5908), 1634. https://doi.org/10.1126/science.1161907

Melore, C. (2022). Family tree stumped: Most Americans can’t name all 4 of their grandparents! Family & Relationships News, April 6, StudyFinds. https://studyfinds.org/family-tree-name-grandparents/

Petr, M., Pääbo, S., Kelso, J., & Vernot, B. (2019). Limits of long-term selection against Neandertal introgression. Proceedings of the National Academy of Sciences, 116(5), 1639-1644. https://doi.org/10.1073/pnas.1814338116  

Sankararaman, S., Mallick, S., Patterson, N., Reich, D., Djebali, S., Tilgner, H., Guernec, G., Martin, D., Merkel, A., Knowles, D.G., et al. (2016). The combined landscape of Denisovan and Neanderthal ancestry in present-day humans. Current Biology, 26(9), 1241-1247. https://doi.org/10.1016/j.cub.2016.03.037

Suntsova, M.V., & Buzdin, A.A. (2020). Differences between human and chimpanzee genomes and their implications in gene expression, protein functions and biochemical properties of the two species. BMC Genomics, 21 (Suppl 7), 535. https://doi.org/10.1186/s12864-020-06962-8