Last year 23andme’s research team published a major landmark study titled “Genetic Consequences of the Transatlantic Slave Trade in the Americas“. Arguably the largest DNA study to examine African ancestry in the Americas! Covering a wide span of the Afro-Diaspora, incl. also several thousands of African Americans. Highly interesting therefore. The research approach of this study consists of combining genetic data obtained from 23andme customers with Trans-Atlantic Slave Trade patterns. Which is practically the same approach I have been using ever since I started my AncestryDNA survey in 2013. This study by 23andme is even including Cape Verdean samples as a control group! Which is something I have done as well in all my research sofar.1 Since I have recently finished my survey findings based on 23andme results (2018/2019 version) it should be useful to compare notes.
23andme’s 2020 study
- Genetic Consequences of the Transatlantic Slave Trade in the Americas (Micheletti et al., 2020)
- Supplemental Data Micheletti et al, 2020
- Table S7. Mean Ancestry Composition for the Research Cohort (main source for comparison)
My own survey findings based on 23andme and AncestryDNA results
- Afro-Diaspora AncestryDNA results: A Comparison (Fonte Felipe, 2016)
- Update: Afro-Diasporan AncestryDNA Survey (part 1) & part 2) (Fonte Felipe, 2018)
- 23andme’s African breakdown put to the test: Afro Diaspora edition! (Fonte Felipe, 2021)
Table 1 (click to enlarge)
In this blogpost I will compare my own research findings (based on regional admixture) with 23andme’s study from 2020. In fact much of the data contained in 23andme’s study (based on the 2018 version of Ancestry Composition) is consistent with my own. As demonstrated above in Table 1. Which features the African breakdown for African Americans on 23andme (scaled to 100%).2 Despite smaller sample size on my part actually very similar outcomes. Providing mutual corroboration. The study’s main findings of lower Senegambian and higher Nigerian ancestry than expected for African Americans are in line with what I had already established in my 2015 survey. Based on AncestryDNA test results for 350 African Americans. As well as more recently in my 23andme survey. See also:
- 350 African American AncestryDNA results (2015)
- 200 African American 23andme results (2018/2019 version)
Within the remaining part of this blog post I will discuss the following:
- Why do so many African Americans have Nigerian ancestry?
- Ancestral implications of “Nigerian” go beyond modern-day borders
- Domestic Slave Trade from mostly Virginia spreading Bight of Biafra lineage
- Lower Senegambian than expected because of less reproduction?
- Substructure according to state origins
- African breakdown for other parts of the Afro-Diaspora
- Mostly in agreement with historical expectations
- Overlap & differences with my own survey findings
- Regional diversity and substructure
- Confirmation of Upper Guinean Founding Effect?
- Discordances & limitations of 23andme’s study
- Afro-descended samples taken from migrants underrepresent wider variation in countries of origin
- Central African IBD disproportionately high when contrasted with regional admixture from Central Africa
- Sex-biased admixture: multiple & context-dependent historical narratives!
- Exciting future prospects: personalized 23andme results featuring African IBD specified according to ethnic groups
1) Why do so many African Americans have Nigerian ancestry?
Ancestral implications of “Nigerian” go beyond modern-day borders
Map 1.1 (click to enlarge)
“As much as two thirds of African-Americans’ Sub-Saharan DNA may trace back to Nigerian ancestors, due to the disproportionate impact of the Atlantic slave trade on the people of the region.” (23andme’s website, 2018)
“Far more people in the US and Latin America have Nigerian ancestry than expected, given what historical records show about the enslaved people that embarked from ports along present-day Nigeria into the Americas, according to the study. What that suggests: This is most likely a reflection of the intercolonial slave trade that occurred largely from the British Caribbean to other parts of the Americas between 1619 and 1807, Micheletti and Mountain wrote.” (CNN, 2020)
“While groups from modern Nigeria appear to loom large in African North American ancestry, as Micheletti and colleagues have indicated, there are specific reasons for and limitations to this observation. Only when armed with the proper historical and cultural context data can we rely upon the genetics and genomics to fathfully reconstruct a people’s population biology”.” (Jackson, 2020, p.206)
“Genetic Consequences of the Transatlantic Slave Trade in the Americas” is based on an unprecedented number of over 50,000 samples (see Figure 3.1). Although often anticipated by earlier studies the data produced by 23andme’s study is therefore highly significant.3 With potentially wider implications! Given also the robustness of 23andme’s genotyping and regional admixture analysis. Which is often seen as one of the best if not the best on the market. Naturally keeping in mind all of its inherent limitations (see also this page). Therefore I certainly think 23andme did an outstanding job by publishing their research findings. I am especially appreciative of the detailed supplemental data. However I do not always agree with the historical reasoning found within 23andme’s study.
This goes in particular for how the authors explain their main finding of “Nigerian” DNA being unexpectedly “overrepresented” among African Americans. This research outcome was also widely copied by press articles released by BBC, CNN, The New York Times etc.. At times uncritically so, which may have lead to misconceptions among the general public. Within 23andme’s study (p.272) this finding is discussed mostly within the context of their IBD analysis (shared Identical By Descent DNA segments, see Figure 3.2). But this finding is also observable when going by regional admixture (see Table 1). The seemingly consistent and usually quite high “Nigerian” scores on 23andme may indeed already have come across as a perhaps overplayed theme for many African Americans.4
And to be sure there are several caveats to be made.5 One of them is shown in Map 1.1 above. Portions of what is being classified as “Nigerian” DNA might actually have been inherited from ancestors from neighbouring countries, such as Benin and Cameroon. But also Togo and even (eastern) Ghana. Combined also forming the macro-region of so-called Lower Guinea.
However I do believe that the greater part of “Nigeran” scores is indeed pinpointing (southern) Nigerian ancestry for African Americans. Based on the relative share of Bight of Benin in Trans-Atlantic slave trade to North America being quite low (around 3%, see this page). While also the documented share of Cameroonian captives is known to have been quite subdued (an estimated 5% of total Bight of Biafra numbers, see also this discussion). Also in other ways a predominantly (southern) Nigerian interpretation of “Nigerian” scores can be corroborated. In particular by the frequency of Nigerian DNA matches for African Americans.
But do also notice from Table 1 that a considerable share of African DNA (~20%) is being left unspecified by 23andme’s analysis under one of its “Broadly” categories. Making for a further qualification of “Nigerian” scores among African Americans. Afterall if it turns out that most of this unspecified part is actually Central African DNA than this would change the overall proportions considerably. The upgrade of 23andme’s algorithm in 2020 has indeed already resulted in somewhat higher “Angolan & Congolese” scores. Although actually also “Nigerian” scores have been on the increase as well. While “Broadly West African” is remaining with a considerable (scaled) share of around 10%-13%. So still some ground to cover by 23andme.
- Comparing 23andme’s 2018 & 2020 version, based on group averages for 2 African American survey groups (n=100 & n=200)
Future updates and the introduction of new West African categories might very well lead to a finer regional resolution of 23andme’s African breakdown. In which “Nigerian” scores will be less conspicious. Although again I greatly suspect that a substantial degree of Nigerian ancestry among African Americans will still be the outcome of improved DNA testing as well. Quite likely somewhere in the range of 20%-35% on average (within an African breakdown scaled to 100%). Future research will be needed to narrow down to a more solid estimate of course.
However it is clearly an exaggeration to claim that as much as 66% of the African breakdown for African Americans is “Nigerian” as seen above in the quotations. This statement can still be seen on 23andme’s website (see also this screenshot). Even when in fact it is not supported at all by 23andme’s own study from 2020 in which instead a scaled group average of around 36% was revealed (see Table 1 which is based on the study’s supplement mmc3/Table S7).
On the other hand the high incidence of “Nigerian” scores among African Americans certainly doesn’t come out of thin air either! Like many of the study’s other findings it could still be a plausible outcome when looking at the relevant context and the relevant statistics!
Domestic Slave Trade from mostly Virginia spreading Bight of Biafra lineage
Figure 1.1 (click to enlarge)
Map 1.2 (click to enlarge)
Table 1.1 (click to enlarge)
Table 1.2 (click to enlarge)
“The Domestic Slave Trade, also known as the Second Middle Passage, adds yet another ugly chapter in the history of American Slavery. But it’s still very important to study because of the farreaching implications for family ties across the USA and also tracing back to ultimately Africa via ancestors from Virginia. […] I suppose only future DNA studies using large scale sampledata might confirm these ancestral connections.” (Fonte Felipe, 2015)
“Virginia is still clearly the greatest source of relocated slaves. In fact in the previous decades especially Kentucky and North Carolina might have depended heavily on Virginia’s slave exports to such an extent that their slave populations could be a very close reflection of Virginia’s African ethnic composition with its prominent connection to the Bight of Biafra. Even South Carolina, the biggest importer of slaves directly from Africa, still also imported many slaves from Virginia in the late 1700’s.” (Fonte Felipe, 2015)
The information shown above is crucial for a deeper understanding of African American demographics and their main African regional roots. Which might often have been inherited by way of Virginia-born ancestors from the early 1800’s as well as the 1700’s. To a greater degree than still imagined among the general public. Incorporating and dispersing a major Bight of Biafra connection (mainly southeast Nigeria, see this page for maps) to be traced back mainly to the early 1700’s (see Figure 1.1). Quite likely already reflected (imperfectly) in primary “Nigerian” scores reported for so many African Americans on 23andme as well as on Ancestry in fact!
As can be seen in Table 1.2 it appears that a majority of African Americans had direct Chesapeake/Virginia origins up till 1810 at least. Due to an early occurrence of cumulative population growth especially in and around Virginia. Subsequent Domestic Slave Trade from mainly Virginia (see Map 1.2 and Table 1.1) undoubtedly resulted in a further spread and dispersal of indirect but still ultimate Virginia/Chesapeake origins within the USA. Even when often obscured by family recollection as well as documented records usually only going back to the late 1800’s. The share of Chesapeake and Inland (VA +MD+ KY +TN) actually might still underestimate Virginia-derived origins as I did not include North Carolina out of caution. Even when this state is known to have been settled and supplied with enslaved labourers by neighbouring Virginia to a great extent as well (see this chart).
Historians have been aware for quite some time already of the immense significance of Domestic Slave Trade and its main driving force hailing from Virginia. Also the prominent share of the Bight of Biafra in direct Trans-Atlantic Slave Trade to Virginia/Maryland has been known since atleast 1969. When Philip D. Curtin published his highly influential “The Atlantic Slave Trade: A Census” (see this overview). Still many people appear to be lacking this vital piece of knowledge.
Or simply haven’t put two and two together yet. Possibly caused by a strict focus on overall Trans-Atlantic Slave Trade patterns for the USA. In which the numbers for South Carolina, the biggest importer of African captives into North America, tend to distort the picture. It is however often forgotten that South Carolina’s enslaved population still was only half the size of Virginia’s around the time Trans Atlantic Slave Trade was officially abolished in 1808 (see Table 1.2). Which is to be explained by decisive differences between Virginia and South Carolina in relation to mortality, reproduction as well as periods of arrival of African captives (see this chart). Resulting in divergent demographic trajectories and a much earlier onset of rapid self-sustaining population growth among Virginians.
Also 23andme’s research team as well as their consultants seem to have been surprised by their finding of a relatively high share of “Nigerian” among African Americans. Because apparently they were under the false belief that “historical records showed little evidence of direct transport between Nigeria and the US” (see this article). Astonishingly even the New York Times seems to have been seriously misinformed. Although to their credit they did place a correction afterwards (see this link).
23andme’s study is barely mentioning Domestic Slave Trade at all. Except when pointing out how this forced movement of people “could easily obscure pre-existing local population structure” (Micheletti et al., 2020, p.274). Instead 23andme’s research team and it consultants identified Intra-American Slave Trade from the British Caribbean into North America as the main reason for the increased level of Nigerian ancestry they encountered among African Americans.
Most likely this additional flow of incoming African captives has been a contributing factor indeed. Aside from introducing additional Nigerian lineage actually also responsible for additional ancestry from especially Ghana as well as from Benin and Togo.6 However as mentioned in 23andme’s own study itself (p.272) this Intra-American Slave Trade was clearly minor and secondary compared to overall Trans-Atlantic Slave trade. According to estimates for the USA it could be around 15% of total slave trade (overseas). And only 10% for Virginia and South Carolina! See also:
- Estimated Slave Imports To North America, 1619-1810 (taken from O’Malley, 2009),
- Slave Voyages: not only Trans-Atlantic but also Intra-American! (Fonte Felipe, 2020)
While actually going by sheer numbers Domestic Slave Trade was easily most significant for the USA. An estimated 1 million enslaved African Americans (often with Virginia background, see Table 1.1) are known to have been victimized by the so-called Second Middle Passage. To be sure I am not pointing this out to diminish the role played by Intra-American Slave Trade! Especially by way of early founding effects I imagine this flow of people might still have been disproportionally impactful for certain states. However overall speaking it seems quite self-evident that Domestic Slave Trade played a major if not primary role as well in shaping the African American genepool and its main regional African roots. Either way certainly not a factor to be disregarded!
The argument linking these forced domestic migrations with increased levels of Nigerian ancestry is most compellingly illustrated by the main Bight of Biafra connection for Virginia. A topic I blogged about already in 2015. Also in relation to my AncestryDNA survey at that time. In fact additional factors may have reinforced or even augmented the genetic impact derived from southeast Nigeria by way of Virginia. In particular a more even gender ratio among captives brought in from the Bight of Biafra as well as cumulative population growth driven by natural increase being more pronounced in Virginia and surrounding states when compared with South Carolina.
“Richmond, Virginia, became a central slave market that facilitated the interstate slave trade. Planters in the Old South chose to breed and sell enslaved Africans and African Americans as a commodity rather than freeing them. This domestic slave trade, […] made enslaved Africans of the Chesapeake Bay the primary targets of this new strategy. The main African ethnic groups of this region were the Ibo and Ibo-related peoples from what is now modern Nigeria. These groups disproportionately bore the brunt of this cruel and inhumane activity. Yet it is their lineages now that are so abundantly represented among African North Americans.” (Jackson, 2020, p.204)
A similar argument has actually also been made by Dr. Fatimah Jackson, who wrote a review article about 23andme’s study in the same journal it was published. See quotation above. She also explicitly mentions “breeding farms” in Virginia being an important factor. In return 23andme’s research team wrote a response to Dr. Fatimah Jackson. To their credit conceding that the additional role played by Domestic Slave Trade and in particular Virginia’s Bight of Biafra connection may indeed have been overlooked in 23andme’s study.7 Naturally further research will be required for greater clarification. I myself intend to follow-up on the possible extent of Virginia’s genetic impact on the overall African Americans genepool. Basing myself also on African DNA matching patterns for African Americans. In the meanwhile for more detailed discussion and references see:
- The Igbo Connection for Virginia & Virginia-Descendants (Fonte Felipe, 2015)
- So many Nigerians: why is Nigeria overrepresented as the ancestral genetic homeland of Legacy African North Americans? (F. Jackson, 2021)
- Response to Jackson (Micheletti et al., 2021)
Lower Senegambian than expected because of less reproduction?
Table 1.3 (click to enlarge)
“When wanting to correlate slave trade statistics with the ancestral origins of currentday populations it seems therefore one should keep in mind additional factors (see also this page). The ethnic demographics of African captives after arrival in the Americas are a big unknown for the most part, especially when multigenerational timeperiods are being considered. Some regional/ethnic groups from Africa might possibly have been better positioned to have more descendants than other groups. For several reasons such as a favourable gender ratio (more females), cumulative growth because of earlier creolization, greater victimization by domestic slave trade, being disproportionally located in areas with less mortality etc., etc. Either way resulting in relatively more DNA markers inherited from these groups to be identified within the current day African American genepool. Inspite of all the limitations of my AncestryDNA analysis it might be that the data I collected is already reflecting some of these demographic processes.” (Fonte Felipe, 2015)
“It might be that the ethnic composition of socalled Senegambians being brought to the USA has been misjudged by historians based on incomplete documentation. With the relative contribution of people located more towards the interior [towards western Mali] and also to the south [towards Sierra Leone] being underestimated.” (Fonte Felipe, 2015)
“Given this growing amount of DNA evidence it seems the assumed prominence of Senegambian ancestry among African Americans is up for reconsideration. Upper Guinea is often proposed as one of the biggest “homelands” for African Americans based on historical documentation. And to be sure African American AncestryDNA results do indeed confirm the widespread existence of these lineages. Still […] Upper Guinean origins on average do appear to be more diluted and subdued than expected. This also includes results from Louisiana for whom the Inter-Colonial and Domestic Slave Trade is often underestimated when considering their roots within Africa.” (Fonte Felipe, 2016)
Aside from Nigerian DNA being “overrepresented” among African Americans 23andme’s study also singled out as one of their main findings that Senegambian ancestry was lower than expected. Again both outcomes were already anticipated and described in my previous AncestryDNA survey of 2015 (scroll down to “Less Upper Guinean ancestry than expected?” or see also this comment). The quotations above still reflect my main reasoning now that my 2015 finding has been replicated by 23andme’s study.8
The baseline being used to determine how high Senegambian ancestry “ought to be” is the share of Senegambia (see this map) in direct Trans-Atlantic slave trade to North America. Which would be somewhere in the 20-25% range.9 However based on 23andme’s study as well as actually my own survey findings “Senegambian & Guinean” represents a share of less than 10% of total African ancestry for African Americans (see Table 1). Which would be about half as much as might be expected. Also going by so-called IBD (based on identical DNA segments rather than regional admixture) 23andme established a lower than expected share from Senegambia (see Figure 3.2). To be sure Senegambian ancestry is still present with a considerable and distinctive share among African Americans!
As discussed in the previous section it is first of all essential to realize that aside from Trans-Atlantic Slave Trade also Intra-American and probably even more so Domestic Slave Trade had a major impact on the African American genepool. Resulting in modified regional African admixture proportions. In 23andme’s study (pp. 272-273) the lower than expected level of Senegambian ancestry among African Americans is explained by 1) “increasing rates of deportation of children from Senegambia over time” 2) “higher mortality due to dangerous plantation conditions“. The last factor further detailed by referring to being exposed to malaria, esp. in the rice plantations of South Carolina. To which Dr. Fatimah Jackson’s response is as follows:
“Jackson argues that a more thorough understanding of the history could offer other explanations. Senegambians, she says, likely weren’t “any more exposed to malaria than anyone else.” Rather, she attributes their reduced genetic presence to two other possibilities: Senegambians were known to be instigators of many notable rebellions, especially during the Middle Passage. The consequence of a failed rebellion, she notes, is death. Alternatively, many Senegambians were Muslim, meaning they were less likely to marry or have children with others outside their religion” (The Scientist, 2020)
I personally believe that multiple causes may explain the greater degree of dilution of Senegambian lineage. All of which are in need of further investigation. The two circumstances mentioned by 23andme’s research team may certainly have been contributing factors. However I am not convinced that they would be among the primary causes. For one thing Senegambian captives were not restricted to just South Carolina. It is often not realized that also Virginia received a relatively high proportion of enslaved Senegambians. In fact only second in number to the people who came from the Bight of Biafra (see Figure 1.1). Also the relative share of Trans-Atlantic slave trade with Senegambia being quite similar to South Carolina in fact (around 20%, see this overview).
Given that many African Americans are likely to have mostly Virginia/Chesapeake derived origins (see Table 1.2) I think that further research into localized creolization in that area should be particularly fruitful. Which is to say the overall context which lead to the first generations of USA-born African Americans. The gender ratio among both Senegambian and Biafran captives should be highly relevant in this regard. Especially given that it seems the average male percentage among Senegambians might even have been higher in Virginia (around 80%, see this overview). A relatively higher number of Biafran women versus a relatively lower number of Senegambian women is bound to have had an effect on their overall reproduction rate. The greater genetic inheritance by way of African women instead of African men is a topic which is actually also explored in 23andme’s study. Albeit from a different perspective (see p. 273: sex-bias discordance). But in addition also cultural factors may have played an important but as of yet understudied role as highlighted by Dr. Fatimah Jackson.10
Substructure according to state origins
Table 1.4 (click to enlarge)
Table 1.5 (click to enlarge)
Map 1.3 (click to enlarge)
“which regional African slave trade patterns have been most impactful on the overall African American genepool: those from Virginia or those from South Carolina? Of course I do realize things might be more complex and other factors may play a role as well. But this could very well offer a (partial) explanation for the high frequency of primary “Nigeria” scores among most African Americans. ” (Fonte Felipe, 2021)
“Learning more about the approximate genetic contribution of either Virginia or South Carolina in other USA states might very well be crucial in obtaining a greater understanding of African Americans as an ethnic group in their own right. Not only in regards to their African origins. But also I imagine in regards to the formation of mainstream African American culture and how it evolved later on. And possibly even how some minor non-African admixture may have originally occurred in the Virginia/Maryland area and later on (due to either forced or voluntary migrations) was dispersed among African Americans nowadays living in other states.” (Fonte Felipe, 2021)
“I suppose this can serve as some sort of confirmation already that South Carolina has higher than average Upper Guinean ancestry as well as from Central Africa.” (Fonte Felipe 2015)
Genetic substructure is basically referring to subgroups within greater populations. Despite commonalities various localized factors may still have caused differentiation between various subgroups within a given population. In particular pointing towards a distinctive mix of African regional origins. Localized substructure might exist whenever slave trade patterns are known to have been markedly different for various entry points within one country. Showing overlap to be sure but still recognizable due to deviating proportions. With proper interpretation this can be very helpful in your quest to Trace African Roots!
This is a theme I have been researching for African Americans already since 2013 when I first started my previous Ancestry survey (n=350). Because of greater detail in Ancestry’s West African breakdown the underlying trends were more clearcut then. But also in my more recent 23andme survey I came across meaningful differentiation.11 Roughly speaking conforming with known differences in Trans-Atlantic slave trade patterns between South Carolina and Virginia in particular. See also:
- African regional origins of Trans-Atlantic Slave Trade for Virginia and South Carolina (source: slavevoyages.org)
- African breakdown according to state origins (VA, SC and LA) (AncestryDNA 2013-2018)
- African breakdown according to state origins (VA, SC and LA) (23andme, 2018/2019 version)
Regrettably within 23andme’s study this topic is not really explored in depth. Although the possibility of substucture is briefly discussed. But it is assumed that multiple migrations (both forced and voluntary ones) “could easily obscure pre-existing local population structure in African-descendant communities as a result of admixture” (see Regional Discordance, p.274). Which certainly seems to be true. Because overall speaking African Americans across the country do indeed display rather similar African breakdowns on 23andme.12 With “Nigerian” generally speaking showing up as most significant component. On the other hand as shown in Table 1.4 and Map 1.3 above minor yet distinctive differences can still be picked up, also based on 23andme’s own data-set.
And with careful analysis I would argue that these variations can already be correlated with historical expectations. As outlined above the genetic impact of a mostly Virginia/Chesapeake derived Domestic Slave Trade appears to be widespread throughout the Deep South. As indicated by the primary level of “Nigerian” across the country. However I find it very telling that “Nigerian” is most elevated for Chesapeake (DE, MD, VA, WV) and East Inland (KY, TN). As this would be in line with Virginia and directly surrounding states having a more pronounced Bight of Biafra connection. Kentucky and Tennessee being among the earliest destinations of Virginian-born African Americans forcibly being branched out to other parts of the country.
Less conspicuous but still noteworthy also how South Atlantic (which includes South Carolina) has a somewhat higher group average for “Ghanaian, Liberian & Sierra Leonean” as well as for “Angolan & Congolese”. Because of the way 23andme’s research team has chosen to group and select their American (4gp) samples the patterns might not yet come out with optimized full effect. For example combining North Carolina with South Carolina & Georgia in “South Atlantic” will tend to downplay the distinctiveness of any South Carolinian derived regional patterns. Because North Carolina arguably has more in common with Virginia’s main African regional roots. Combining Louisiana with Texas will also result in a loss of any potential zooming into particular features for Louisiana. See also Table 1.5.
Actually 23andme’s American (4gp) samples might not per se be self-identified African Americans. But rather either biracials or people with multigenerational migrant origins. Afro-descended indeed (atleast > 5%) but without USA-born lineage on all sides to be traced back beyond the Civil War. Possibly including either Hispanic Americans or even Cape Verdean-Americans as well. I greatly suspect that this goes especially for 23andme’s data for the Northern States and Midwest (judging from atypical levels of non-African admixture).13 So we have to be careful in drawing any firm conclusions based on Table 1.4. Atleast not in all regards. On the other hand Map 1.3 is also quite illustrative of the informational potential of a substructure approach. Highlighting how especially certain areas within South Carolina (Lowcountry?) appear to have a greater pull towards Upper Guinea (also incl. Sierra Leone and Liberia!) as well as Central African DNA.
Table 1.6 (click to enlarge)
Judging from my most recent surveyfindings, based on updated 23andme results (2020 version), it seems that greater as well as more insightful differentiation can indeed be established. Especially when making sure of a careful selection of sample groups. Not only between states but actually also within states, as demonstrated for South Carolina in Table 1.7 below. Particularly insightful when wanting to grasp the localized formation of the Gullah people in the Lowcountry and adjacent Pee Dee area. Take notice that on average my South Carolinian survey group (n=100) does not show “Nigerian” as primary African region, but instead “Ghanaian, Liberian and Sierra Leonean”! (see also this overview for frequency of primary African regions).
Another very useful aspect suggested by Table 1.6 is that the relative impact of either South Carolina or Virginia derived African origins can be deduced from looking into results of neighbouring states. Georgia and Florida being more similar to South Carolina. As indicated especially by above average scores for “Senegambian & Guinean” and “Angolan & Congolese”. As well as a greater frequency of primary “Ghanaian, Liberian & Sierra Leonean” scores (see this overview). Probably indicative mostly of Rice Coast lineage, although additional Ghanaian DNA will also be included to varying degree.
While North Carolina is clearly tilting towards Virginia and its main Bight of Biafra origins. Especially noticeable from its elevated “Nigerian” level. My Louisiana survey group (incl. 5 most likely Creole persons) is showing more similar patterns to South Carolina. Although probably for independent reasons. Intriguingly next-door state Mississippi is much more so following the prevailing “Nigerian” trend. As most clearly demonstrated and probably also mostly originating from Virginia. For more detailed discussion:
Table 1.7 (click to enlarge)
2) African breakdown for other parts of the Afro-Diaspora
Mostly in agreement with historical expectations
Figure 2.1 (click to enlarge)
Table 2.1 (click to enlarge)
“Here, we analyzed genotype array data from 50,281 research participants, which—combined with historical shipping documents—illustrate that the current genetic landscape of the Americas is largely concordant with expectations derived from documentation of slave voyages. […] While some discordances can be explained by additional records of deportations within the Americas, other discordances yield insights into variable survival rates and timing of arrival of enslaved people from specific regions of Africa.” (Micheletti et al., 2020, p.265)
“[…] most of the findings […] indeed seem to be in line with the documented African roots for my sample groups, even if obviously not exactly so. Other outcomes might seem more unexpected but with additional reasoning these findings also make sense historically speaking or prove to be valuable for an improved interpretation […] ” (Fonte Felipe, 2016)
“Also my survey of Afro-Diaspora results has largely been a confirmation of historically documented African origins for each nationality. […]. Again such potentially profound information is not something to carelessly brush aside when wanting to Trace African Roots!” (Fonte Felipe, 2018)
In this section I will discuss the regional admixture findings of 23andme’s study for other parts of the Afro-Diaspora. In fact again greatly anticipated by my previous survey findings on Ancestry (2013-2018) as well as remarkably conforming with my more recent 23andme research outcomes (2018-2020). Overall sample size being the main difference. But when it comes to coverage actually already in 2015 I had analyzed AncestryDNA results from several parts of the Afro-Diaspora (in addition to the USA: Jamaica, Haiti, Puerto Rico, Dominican Republic, Mexico and Cape Verde). Gradually extending my survey with Brazil in 2018 and many other countries. It will be useful therefore to consult my previous blog posts dealing with these findings. As I will attempt not to repeat myself while discussing the usually very similar results from 23andme’s study. See also:
- Documented African origins confirmed on group level? (scroll down to section 4) (Fonte Felipe, 2016)
- African breakdown on Ancestry for various parts of the Afro-Diaspora (Fonte Felipe, 2018)
- Historical plausibility for 23andme results (last part of section 1) (Fonte Felipe, 2021)
The extraordinary dataset produced by 23andme’s 2020 study (mainly contained in Table S7) is replicating and corroborating my earlier discovery in 2015 that when it comes to regional admixture DNA testing for Afro-descendants is able to provide historically plausible results. Atleast roughly so and also taking into account the additional impact of Domestic and Intra-American Slave Trade or post-Slavery migrations, whenever relevant. Something which I have described in fact not only for African Americans but also for Haitians and Hispanic Americans already in 2015. Of course each DNA testing company should be judged on its own merits. As there will often be variation in performance. Also after each update of their estimates 😉 However when based on 23andme results as well as previously based on AncestryDNA results this major observation seems to hold true alright.
Naturally there is still plenty of room for improvement and further specification. Therefore some sceptics might say what’s the big deal? But given all the uncertainties & unknowns Afro-descendants have to face when wanting to Trace African Roots this outcome is in fact highly significant and encouraging! As it implies that DNA testing can indeed be a valuable building stone for follow-up research. And is not “just for entertainment“. Unrealistic expectations about “100% accuracy” as well as counter-productive obsessing about regional labeling should be avoided of course. Instead focus on what ever informational value you can obtain despite imperfections (see also footnote 5).
Because of Ancestry’s more detailed West African breakdown it was at times easier to make the link with historical expectations during my previous Ancestry survey. This goes especially for singling out the prominent Bight of Benin connections for Haitians and Brazilians (Bahia) by way of “Benin/Togo”. While the so-called “Mali” category was helpful in providing a broader concept of Upper Guinean DNA. In fact for African Americans the combined share of “Senegal” and “Mali” (around 17%) was suggestive of a lesser degree of dilution of Upper Guinean ancestry than discussed in section 1, based on the average level of “Senegambian & Guinean”.
On the other hand 23andme’s African breakdown arguably has a much better proxy for Southeast African DNA by way of its “Southern East African” category (listed under its old name “Eastern Bantu” within the supplemental data). The regional framework for West Africa might be more basic on 23andme. But it is still very useful for signaling certain key characteristics in African regional roots for different parts of the Afro-Diaspora. Also in other ways I find that 23andme is usually producing more consistent and coherent admixture estimates than Ancestry’s 2013-2018 version. The additional non-African scores and Recent Ancestral Locations usually reinforcing the robustness of 23andme’s predictions.
Figure 2.1 features the main findings of 23andme’ study in relation to regional admixture. Based on the summarized 23andme results for an amazing number of 27,422 persons with atleast 5% African admixture from allover the Americas as well as Cape Verde! I have translated the visuals into actual percentages in Table 2.1 (by scaling the African breakdown found in the study’s supplemental data Table S7). Take note how “Senegambian” is clearly peaking for Cape Verdeans. And this category is also in primary place for various Hispanic American populations. “Coastal West African”, a.k.a. ”Ghanaian, Liberian & Sierra Leonean”, is widespread but only most significant for Guyana/Suriname as well as the Bahamas.
“Nigerian” is mostly prevailing for the USA and the British/French Caribbean. Again quite widespread but “Nigerian” is certainly not a default outcome when taking into account the entire Afro-Diapora! When comparing with other groups such as Jamaicans you get a greater sense of the variability and how the USA fits in the greater picture. As actually “Nigerian” is far from being predominant for African Americans. Especially ”Ghanaian, Liberian & Sierra Leonean” also being substantial. “Angolan & Congolese”, (formerly just “Congolese” which was a mere name change) quite tellingly is showing up as biggest component in only a few places: Brazil, Argentina and Uruguay. Even more evocative to see how the usually quite subdued “Southern East African” category is also clearly peaking in these countries, especially Rio de La Plata. As a substantial Mozambique connection for this area is very likely.
These patterns are mostly in agreement with historical expectations and published DNA studies. As I have discussed in greater detail elsewhere. Central & Southeast African scores are already quite distinctive and in the expected ranking order for most survey groups (notice for example the relatively high share of 15.1% “Angolan & Congolese” for Haitians). However this mainly Central African component is most likely underestimated in 23andme’s 2018/2019 version. Possibly to some extent hidden under the usually considerable “Broadly” scores (13%-41%). This can be most easily verified by contrasting with my previous Ancestry survey findings, my African 23andme surveyfindings as well as by comparing with Trans-Atlantic Slave Trade patterns.
Fortunately it seems that after the 2020 update 23andme has improved its detection of Central African DNA. Also very hopeful to learn that actual Angolan samples are now available to 23andme (presently only used for IBD, see Table 3.1). Then again there is still some ground to cover. In particular given that ambivalent “Broadly West African” scores are still remaining at a considerable level. See also:
- Update of 23andme’s African breakdown (Fonte Felipe, 2020)
Overlap & differences with my own survey findings
- Overall research outcomes corresponding with historical plausibility
- Scaled group averages very similar for the most part
- Shared coverage across the wider Afro-Diaspora (incl. Cape Verde!), except for Rio de la Plata & Indian Oceanic Afro-Diaspora
- Larger sample size within 23andme’s study
- Substructure within Afro-descended populations only specified in my survey findings
- Regional diversity highest among Hispanic Caribbeans according to my analysis
- Contrasting historical explanations, esp. in regards with African Americans and Hispanic Americans
Table 2.2 (click to enlarge)
“Establishing where each African region is relatively more pronounced or instead more subdued might provide insightful clues for the unique ethnogenesis of each nationality being shown as well as for African Americans.” (Fonte Felipe 2015)
“Unlike commonly assumed you do not need to sample entire populations to obtain informational value with wider implications. Naturally greater sample size does (usually) help matters.” (Fonte Felipe, 2018)
- Distribution of 23andme’s 50,281 research participants, incl. 27,422 from the Americas & Cape Verde
- Afro-descended nationalities featured in Micheletti. et al. (2020) (Table S2 in the supplement)
I have already discussed above how 23andme’s study from 2020 and my own 23andme surveys are greatly similar in our research approach. Sharing the main outcome that 23andme’s test results for Afro-descendants are often in line with historical plausibility. Albeit that I do not always agree with the specific historical explanations given by 23andme’s researchteam to account for the discordances they came across. The overview above is featuring a direct comparison between my own findings for various survey groups from across the Afro-Diaspora and the corresponding ones (bolded) from 23andme’s study (based on the 2018 version of Ancestry Composition).
The scaled group averages on display are remarkably close for the most part. For several surveygroups at times an uncannily close similarity was obtained. The minimal differences in some instances being smaller than 1% even! This goes especially for African Americans, Cape Verdeans, Dominicans and Haitians. Providing solid mutual corroboration. But also for the other groups the variation is minor and usually lower than 5%. The greater deviations have been encircled in red and seem to point at sampling limitations on both parts. As well as highlighting wider issues at stake. In particular a most likely underrepresentation of research participants with predominant African ancestry. As I will discuss further below.
Obviously the main difference lies in the samplesize, n=27,422 for 23andme’s study vs. n=889 for my own research. Still given the great similarities shown in Table 2.2 it seems that my more modestly-sized research has been sufficiently robust to pick up on the main tendencies. Unlike commonly assumed you do not need to sample entire populations to obtain informational value with wider implications. Naturally greater sample size does (usually) help matters. However I find it reassuring that also in my previous survey efforts published studies based on (much) larger sample size have usually vindicated or confirmed my own findings. While due to free format on my blog I am often able to provide greater detail and arguably more appropriate context.14
For example whereas 23andme’s study has left out the various “Broadly” categories within their African breakdown. I myself have included them and put them to greater use by also providing an additional macro-regional format. Adding “Broadly West African” scores to the subtotal of West African ancestry and “Broadly Central & Southern East African” into the subtotal of Central & Southeast African ancestry. More basic than the actual African breakdown on 23andme (featuring atleast 12 categories). But I find such an approach to be potentially quite insightful as it enables an intermediate perspective which is often easier to make sense of.15
Also in other ways I have attempted to make my data-set more accessible for wider analysis. Making it easier also to see how individual results fit in the greater picture. By scaling the African breakdown to 100%; providing more detailed continental admixture statistics as well as distribution histograms of total African admixture (e.g. see this one for African Americans). Naturally individual variation is a given and is not to be denied! Any meaningful deviations from the group averages hopefully serving as useful clues.
Despite the big difference in sample size the coverage is actually very similar. In my own survey I have included 25 out of the 37 Afro-descended sample groups found within 23andme’s study (see Table S2). Most of the nationalities missing within my survey being located in either the Caribbean or Central America. But these regions are still represented in my research. As far as I am concerned the most striking novelty within 23andme’s overall sample group is the singular appearance of 29 results from Argentina & Uruguay (“Rio de La Plata”). Not very numerous yet (so apparently also hard to get by for 23andme). But extremely interesting because especially for Uruguay (Montevideo) documented Trans-Atlantic Slave Trade from Southeast Africa had an exceptionally high relative share (see Table 1.33 on this page). And intriguingly this is also reflected in their 14.6% group average (scaled) for “Southern East Africa” (see Table 2.1). Easily the highest in the Atlantic Afro-Diaspora.
But as demonstrated in my own survey “Southern East Africa” is of course even more so a signature region for both South African Coloureds and Indian Ocean Creoles. Not featured in 23andme’s study but the Indian Oceanic Afro-Diaspora was already included in my 2018 survey based on Ancestry results (see this overview). As well as more recently based on 23andme results (see this overview). Looking into their results as well is greatly beneficial for evaluating the overall coherence of 23andme’s African breakdown in my opinion. Plus it also enables a broader conceptualization of the Afro-Diaspora. Trans-Atlantic and Indian Ocean slave trading circuits overlapping precisely in Southeast Africa.
It is regrettable therefore that 23andme’s study does not really explore Southeast African ancestry (beyond regional admixture estimates for “Eastern Bantu” in the supplement). The authors do explicitly mention that this is mainly “due to limited sample representation in Mozambique and Madagascar” (Micheletti et al., 2020, p.266). On the other hand I suppose also by way of including Southeast Asian admixture statistics as well as seemingly Southeast Asian haplogroups more clarity could be given on the seemingly widespread Madagascar connection among African Americans. In my own surveyfindings I did analyze usually minor but still distinctive Southeast Asian admixture scores across the Afro-Diaspora. And I found that these seem to stand out for African Americans (see “Madagascar Connection” on this page).16
Regional diversity and substructure
“From the 1500’s onwards these Spanish colonies were being visited by slave suppliers of numerous European nationalities (Portuguese, Dutch, English, French and even Danes). In this way geographically wideranging circuits of African slave trade were connected to the Hispanic Caribbean area. Quite likely during a longer period of time (especially when considering the clandestine slave trade to Puerto Rico and Cuba in the 1800’s) and including more African slave ports overall than any other destinations in the Americas […]. Seen from that perspective a wider range of African origins for Hispanic Caribbeans and also occurring in more evenly balanced proportions is perhaps to be expected.” (Fonte Felipe, 2016)
“The majority of enslaved people arriving in Latin America (Latin Caribbean, Central America, N. South America, and C. South America) were generally deported from one or two slave trading regions, whereas enslaved people arriving in the United States and British Caribbean were taken from all regions of Atlantic Africa” (Micheletti et al., 2020, p. 271)
The statements above highlight one of the main disagreements between my analysis and 23andme’s study. In my previous Ancestry survey (2013-2018) I already looked into the question of which part of the Afro-Diaspora might have the most regionally diverse African origins. And alternatively which group would be relatively most homogenous in their African regional roots. Using the relative shares of the two main African regions as a rough measure (“∑ Top 2”). As well as looking into the macro-regional breakdown. My preliminary findings in 2016 being that Cape Verdeans (as expected) were greatly uniform in their African origins, overwhelmingly hailing from Upper Guinea. An outcome which has been replicated in all my subsequent surveys. Mexicans and Jamaicans also being relatively homogenous in their African breakdown. A clear focus on Upper Guinean & Central African DNA for Mexicans. And much more so Lower Guinean DNA being predominant for Jamaicans. Again in line with historical expectations.
Unlike 23andme’s study I identified Hispanic Caribbeans as having the most regionally diverse African breakdown. Something which again I was able to correlate with the relevant historical context of both Trans-Atlantic and Intra-American Slave Trade. As well as to be explained by substructure according to level of African admixture. On the other hand 23andme’s researchteam single out African Americans as well as British & French Caribbeans as having genetic connections to a greater number of African regions, on average. Based on both regional admixture as well as IBD. Using the historical justification quoted above (Micheletti, et al., 2020, pp. 270-271). But that statement is frankly speaking questionable or too much of a generalization. Because afterall Intra-American Slave Trade (generally introducing greater variety in African lineage) is known to have impacted the Hispanic Caribbean as well as Panama, Colombia & Venezuela to a large extent (more so than the USA!). And especially when it comes to Puerto Rico and Cuba also for a longer period in time than elsewhere. For more details:
- Dominicans most evenly mixed in their African regions (scroll down to section 3) (Fonte Felipe, 2016)
- Trans-Atlantic and Intra-American Slave Trade for the Hispanic Americas (Slave Voyages Database, 2020)
- Patterns in the intercolonial slave trade across the Americas before the nineteenth century (G. E. O’Malley & A. Borucki, 2017)
In my 23andme survey I have actually not explicitly dealt with this topic of regional diversity. However I suppose a rough measure of regional diversity in more or less balanced proportions could be defined as followed: the number of double-digit group averages for the main 5 African regions (African breakdown being scaled to 100%). Basing myself on Table 2.2 and 23andme’s own data-set it is again Dominicans, as well as Puerto Ricans, Cubans and Colombians who stand out for having an evenly balanced African breakdown. Featuring 4 African regions with group averages greater than 10%. While for African Americans this goes for only 2 such regions (“Nigerian” and “Ghanaian, Liberian & Sierra Leonean”). Cape Verdeans once more standing out for having one clearly predominant African region (nearly 75% “Senegambian & Guinean”). But also Jamaicans, Leeward and Windward Caribbeans as well as Mexicans being relatively homogenous in their African origins. Featuring one clearly major (>40%) primary region.
These divergent research findings probably require further research/interpretation. Keeping in mind also that in my previous analysis of AncestryDNA results I had not yet integrated samples from Cuba, Colombia and Venezuela. While 23andme’s 2020 study has not yet delved into deeper substructure within Latin American populations. As I will discuss in more detail below.
Table 2.3 (click to enlarge)
In section 1 I already described how looking into genetic substructure can be highly beneficial for obtaining greater insight. Basically zooming into a potentially distinctive mix of African regional origins for certain subgroups within a given population. To be defined along geographical, social, cultural, or even “racial” lines. This is a theme I have been researching for African Americans already since 2013 when I first started my previous Ancestry survey. Also for Puerto Ricans I described meaningful substructure in 2015 already. In 2018 I also described substructure for Haitians and Dominicans.
In my more recent 23andme survey I have mostly replicated or even solidified my earlier findings for these groups. Furthermore I have been able to uncover potentially insightful substructure for a wider array of my survey groups. Although often based on an admittedly minimal number of samples. But still usually supported by additional reasoning. Resulting in plausible outcomes when looking at the relevant context and the relevant statistics. This includes Brazilians, Mexicans, Colombians, Cubans and Central Americans (Garifuna!). See section 2 of this page for more details.
- 23andme’s African breakdown put to the test: Afro Diaspora edition! (Fonte Felipe, 2021)
Obviously there might be several factors which could explain genetic results being disproportionate to what you might expect based on slave trade data. But I greatly suspect that in particular substructure within any given Afro-descended population will often also be a highly important factor. Which is why it is unfortunate that generally speaking 23andme’s study does not use this approach of zooming into the sub-national level. Except for African Americans. But as discussed in section 1 their potentially quite useful findings for regions such as Chesapeake and South Atlantic are not really explored in-depth.
However by not zooming into for example the distinctive African regional scores found among “black” or “Afro-Cubans“. While (due to availablity) only mentioning Cuban group averages which heavily gravitate towards African regional patterns for “white” Cubans with minimal African admixture of 5%-15%. You then run the risk of giving a distorted picture of the overall situation. While also potentially drawing premature and unfounded conclusions. The same reasoning can be applied also for Colombia which has a large black minority, especially living in the Pacific region. Judging from my preliminary findings shown below with again distinctive African regional roots when compared with the more mainstream Colombians from esp. the Andes region.
Even more so for Central America it is essential to grasp that on the one hand African regional patterns for in particular El Salvador and Guatemala are greatly similar to the ones for Mexico (focused on Upper Guinea and Central Africa). However for Central American countries with significant Caribbean influence a much greater Lower Guinean impact is inevitable given the presence of the Garifuna in Belize and Honduras as well as numerous descendants of West Indian migrants in Costa Rica and Panama. That last country also being a major transit area for Intra-American Slave Trade conducted by the British into Pacific South America.
Which is why personally I would never have put all of Central America in just one group. Let alone include Belize which is a former British colony! As this is bound to obscure greater underlying variation. As can be seen in Table 2.2. Showing greatly contrasting group averages between my Garifuna and El Salvadorian survey participants. While the summarized data for the Central American sample group from 23andme’s study is not enabling such insight. Certainly an impressive sample size but otherwise undifferentiated and therefore at risk of overgeneralizing things.
Confirmation of Upper Guinean Founding Effect?
Table 2.4 (click to enlarge)
“[…] African Americans have a greater share of Senegambian ancestry only when compared to the English speaking West Indies and Haiti but not so when compared with the Hispanic Caribbean and Mexico/Central America. The persistent Upper Guinean genetic imprint among many Hispanics […] can no longer be ignored“ (Fonte Felipe, 2016)
“almost no captives from Senegambia disembarked directly in either northern South America or Central America after 1650, consistent with our finding of an older TMRCA between Senegambia and these two regions, and the likelihood that earlier African disembarkation resulted in smaller IBD segments between Africans and [Hispanic] Americans of the present day” (Micheletti et al., 2020, p.272)
The widespread frequency of high “Senegambian & Guinean” scores reported by 23andme for many Hispanic Americans is something which was already forecasted to happen. Based on my previous Ancestry survey (2013-2018). An overly USA-centric perspective may have prevented a full realization of how significant Upper Guinean ancestry turns out to be for many Hispanic Americans. Especially in comparison with African Americans. The recent inclusion of early Iberian (Portuguese/Spanish) Slave Voyages into the standard reference Slave Voyages database has been incredibly useful therefore for greater understanding.17
As often these “Senegambian & Guinean” scores seem to reflect the genetic legacy from the earliest victims of the Trans Atlantic Slave Trade (1500’s/1600’s). Many of these persons arrived in the Hispanic Americas from the Upper Guinea region (Senegal, Gambia, Guinea Bissau/Conakry and Sierra Leone), often by way of Cape Verde. I have blogged about this topic many times already (starting in 2014). And I intend to do so again eventually as both my 23andme surveyfindings as well as 23andme’s study outcomes are in support of this remarkable phenomenon!
Returning to Table 2.1 this is plain to see. As only Cuba is not showing “Senegambian & Guinean” as biggest region within the African breakdown (on average!). Instead it is “Nigerian” which is most significant. In line with the known circumstance that Cuba’s slave imports disproportionately took place during the 1800’s. Which is quite atypical for the Hispanic context otherwise (see this overview). Even when the “Senegambian & Guinean” level is still considerable as well (15.4% when scaled). While Table 2.4 is very instrumental in underlining how substructure among certain subsegments of Hispanic populations can be even more so clarifying in understanding the distribution and variation of “Senegambian & Guinean” scores within Hispanic populations.
In 23andme’s study this topic is mainly analyzed not by way of regional admixture but rather by way of their additional IBD findings (based on DNA matching strength).18 This part of 23andme’s study should also be potentially useful. As shown in the quotation above it was established that the shared DNA segments between Hispanic and Senegambian research participants were generally speaking smaller sized and assumed to be older (TMRCA = Time to Most Recent Common Ancestor) than other types of shared African ancestry. Highly suggestive of great dilution. Which would be in line with Upper Guinean ancestry to be traced back all the way to the 1500’s/1600’s. In agreement therefore with my Upper Guinean Founding Effect theory. However I am not convinced that the current IBD data-set from 23andme’s study is sufficiently robust to make any further inferences yet. See also next section.
3) Discordances & limitations of 23andme’s study
“Even though Micheletti et al. provide the largest genetic investigation of the transatlantic slave trade, the study is not without limitations. However, as described above, the discordances explained in Micheletti et al., such as over-representation of Nigerian ancestry, are likely due to events following transatlantic voyages, as opposed to study design.” (Micheletti et al., 2021, p. 210)
“In light of this, the additional historical context provided by Jackson highlights the need for further interdisciplinary research that investigates factors that may have led to differential mortality in enslaved populations.” (Micheletti et al., 2021, p. 210)
“Although the work is commendable for making use of both historical and genetic data, Dr. Alondra Nelson said, it was also “a missed opportunity to take the full step and really collaborate with historians.” (New York Times, 2020)
Up till now I have been focused on comparing the regional admixture results featured in 23andme’s study with my own survey findings. Despite inherent limitations I find the corresponding dataset (mostly contained in Table S7 from supplemental data; as well as featured in Figure 2.1) to be the most valuable part of the study. However 23andme’s research team has actually also utilized other means of investigation. In particular IBD and haplogroups. Certainly potentially useful. However generally speaking I find that this part of the study is most in need of further refinement. Especially the IBD dataset seems to have produced quite a few inconsistencies and discordances. Which to the credit of 23andme’s research team are fully acknowledged and listed at the end of the study. While also an attempt is made to provide possible historical explanations for these deviating outcomes.
These historical explanations however often appear to be based on underdeveloped or haphazard argumentation. Assigning broad generalizations to at times still preliminary research outcomes. And even arguably drawing farfetched conclusions. Naturally with this type of research simplifications need to be made. However I find it truly unfortunate that due to the widespread media attention for this study misconceptions among the general public are being perpetuated. Which goes exactly against the stated intentions of 23andme’s researchteam to “gain further insight into the slave trade“.
Afro-descended samples taken from migrants underrepresent wider variation in countries of origin
Figure 3.1 (click to enlarge)
- Afro-descended nationalities featured in Micheletti. et al. (2020) (Table S2 in the supplement)
“This is likely due to the nonrandom representation of global populations in our study cohort, which consists primarily of US customers and does not target specific admixed populations in Latin America. Although our estimates are lower (about 5%–10% on average), the trend of lower African ancestry in Latin America is still supported by other research.” (Micheletti et al., 2020, p.273)
As admitted by 23andme’s researchteam (see quotation above) their sampling outside of the USA is not per se representative. Despite the study’s impressive sample size for most populations it is crucial to understand that people who migrate to the USA tend to hail from certain overrepresented areas within a given country, due to chain migration. While also going by other socio-economic as well as ethno-racial characteristics they will not be fully representative of the wider variation to be found in their countries of origins. Especially marginalized people, often with greater degree of African ancestry, will therefore be much less in focus. This constitutes a major shortcoming of 23andme’s study. Although to be fair such sampling bias is a very common impediment for DNA studies in general.
Nonetheless I do think that a dedicated sampling strategy could have produced greater insight. In section 2 of this blogpost I already pointed out how the average African admixture scores for my own survey groups often tend to be higher than the corresponding ones from 23andme’s study (see Table 2.2). Notwithstanding my sample size often being relatively small I did still make an extra effort as well to collect as many “Afro-Latino” samples as possible. While also in other ways I tried to collect and compose sample groups which do justice to the full range of diversity for any given nationality. For example also finding samples from northern Haiti, Pacific Colombia or from Bahia state in Brazil. All on a best effort basis naturally because again due to the nature of 23andme’s customer database such samples are often simply not available.
A more thoughtful definition of sample groups would have been beneficial as well. Enabling substructure to be defined on a sub-national level! As I did apply with good effect already in my own surveys. The sample size for Haiti within 23andme’s study is quite impressive for example (n=596). However for a focused analysis of their African roots naturally it would have been preferable not to combine Haitian samples with the ones from the British Caribbean. As was done in 23andme’s study (see Table S2). But instead comparing for example southern Haitians with northern Haitians could have been very insightful (see this preliminary attempt).
Also for a greater understanding of the differentiated African regional patterns in Central America it might have been wiser to leave out Belize all together (Jamaica being much more similar when it comes to African regional roots). As well as form separate subgroups for Panama, Costa Rica on the one hand and El Salvador and Guatemala on the other. Nicaragua and Honduras probably being intermediate. And a most useful distinction then would be according to either Pacific or Caribbean areas. Or better yet provide targeted analysis for small yet distinctive parts of the Afro-Diaspora! Such as I did for my Garifuna survey group.
Another issue concerns the inclusion of people of possibly recent migrant origins within the USA sample group featured within 23andme’s study. As I already mentioned earlier (see footnotes 2 & 13) 23andme’s American (4gp) samples might not per se be self-identified African Americans. But rather either biracials or people with multigenerational migrant origins. Afro-descended indeed (atleast African > 5%) but without USA-born lineage on all sides to be traced back beyond the Civil War. Quite unfortunate as I suppose with adequate quality control this kind of “data contaminination” could have been prevented. But possibly arising from confusion on how to define African Americans an an ethnic group versus “Black Americans” or just Afro-descended Americans in an even broader sense.
I suppose for future studies preventing this sort of sampling error should be part of the standard procedure. Especially when aiming to elucidate ancestral narratives for African Americans. Given the general demographic tendencies in the USA also a more thoughtful stance should be taken then in regards to the ethnic affiliation of African Americans with USA-born lineage on all sides to be traced back beyond the Civil War.
Central African IBD disproportionately high when contrasted with regional admixture from Central Africa
Figure 3.2 (click to enlarge)
Table 3.1 (click to enlarge)
“Local ancestry estimates across the Americas (except for central South America) indicate a lower mean proportion of Congolese (West Central African) ancestry than expected (Figure 3), considering West Central Africa shares more msIBD with the Americas than does any other region in Atlantic Africa (Figure 2A).” (Micheletti et al., 2020, p.274)
One of the main inconsistencies of 23andme’s study is relating to the lower than expected degree of Central African admixture they found when compared with their IBD results. Probably to be explained by several factors. As I have already mentioned before 23andme’s “Angolan & Congolese” category indeed appears to have been underestimating Central African DNA. Although generally speaking the outcomes do atleast provide a credible minimum of such lineage. And these scores are also still in line with historical expectations (based on ranking patterns, see Table 2.1). However I greatly suspect that also the composition of 23andme’s African reference panel is playing a significant role in the IBD outcomes obtained within the study.
As shown in Table 3.1 above 23andme’s African reference set (n=1965) seems to be carefully chosen and is quite representative already of Atlantic Africa. A praiseworthy effort has been made to include historically relevant samples. In particular also including an extra set of Angolan samples. Generally speaking I find it very refreshing to see more attention being given to the significant impact of Central African ancestry across the Afro Diaspora. Which is mostly hailing from Angola and the Congo. Because often this seems to be obscured or misrepresented (e.g. see “Cameroon/Congo” = moreso Angola/Congo for Diasporans?).
On the other hand it is too much of a generalization to assume Central African ancestry will be most significant for practically all parts of the Atlantic Afro-Diaspora! As seems to be implied by 23andme’s IBD findings. The aggregated Trans-Atlantic Slave Trade data (heavily skewed towards Brazil) of course do need to be dissected for each separate disembarkation region. While also subsequent Intra-American Slave Trade and other (forced) migrations need to be accounted for. In my previous AncestryDNA survey (2013-2018) one of my main findings was that Central African DNA was less prevailing than West African DNA, except for Brazil (see this overview). However even in that country certain regions (such as northern Brazil and Bahia) are not likely to have predominant Central African ancestry, based on diverging slave trade patterns. For more details:
Looking more closely into the relative composition Table 3.1 seems imbalanced by a greater and more varied selection of Central African samples. Take note that more than 40% of the samples are from Central Africa (856/1965). While these samples are also easily the most varied ones. Including no less than 17 separate ethnic groups! While for example Senegambia is only represented by one single ethnic group. Although it is stated within the study (p.268) that a correction was applied for imbalances in sample size ((msIBD) I am not convinced that this has resolved things completely. I can easily imagine for example that this sampling imbalance might also explain the lower than expected IBD for Senegambia. As afterall a wider selection of Upper Guinean ethnic groups (incl. Fula) will increase the likelihood of IBD matches, all other things being the same.19
“[…] However I strongly suspect that the frequency of DNA matches from a certain place/ethnic group may not always correlate with autosomal contribution, proportionally speaking. In other words just because my Fula and Malagasy survey participants seem to be extra “matchy” with African Americans does not right away imply that the Fula & the Malagasy represent the biggest ancestral components for African Americans (to be verified by admixture analysis). For the Igbo such a case could arguably already be made. Based on both historical and genetic evidence. But especially for the Malagasy a more subdued & diluted overall ancestral share is to be expected (even if still significant and clearly detectable). Certainly lower than for example the Congolese or Angolan input in the average African breakdown for African Americans.” (Fonte Felipe, 2018)
Although potentially very useful I do think that the current IBD data-set from 23andme’s study is not sufficiently robust yet to serve as a basis for any conclusive statements. With a further improvement of 23andme’s African reference dataset as well as further tweaking of other aspects I suspect 23andme’s IBD findings are likely to change quite a bit. I might still discuss the study’s more extensive IBD findings (incl. also IBD sharing with Europeans as well as Cross-Diapora, see supplemental data!) in future blog posts though.
But for now I would just like to address one more possibly complicating issue. Which is that any over-reliance on DNA matches (IBD) might lead to a disproportional outlook on your complete ancestry. Some ancestral lines being more “matchy” than others as a result of an ethnically skewed customer database. And this may then be corrected by your admixture results. This is something I learnt while researching African DNA matching patterns for Afro-descendants on both 23andme and Ancestry. Perhaps most easily illustrated by how for example a mere 1% European admixture may result in thousands of associated DNA matches. At times also with high amounts of shared DNA. While African DNA matches will usually be much more difficult to detect and will often be small sized. Even for a person who is more than 90% African!
I find it very intriguing how a previous exploration by myself of IBD sharing (or proxy thereof) between Africans and African Americans in 2015 gave rather different results than obtained in 23andme’s study. According to my findings Fula, Igbo and Malagasy testers showed the highest degree of IBD with African Americans. See also quotation above and references below. My research then was based on 23andme’s Countries of Ancestry (CoA) tool when it was still available. Again my samplesize was rather minimal but still with quite likely wider implications. Possibly to be confirmed eventually by Big Data 😉 Either way any information obtained from DNA matches (IBD) is always in need of careful interpretation. While awareness of inherent limitations will be crucial to prevent jumping to misleading conclusions. As is true for any other aspect of DNA testing as well of course. For more details/discussion:
- DNA matches reported by 23andme for 75 Africans (Fonte Felipe, 2018)
- African DNA Matching Patterns (Fonte Felipe, 2017 ongoing)
Sex-biased admixture: multiple & context-dependent historical narratives!
Figure 3.3 (click to enlarge)
“bias towards African female contributions to the gene pools of the Americas due to generations of rape and exploitation. The study found that this was particularly stark in Central America and South America.” (23andme’s website, 2020)
“[…] the significant differences between the U.S. and Latin America was a surprise. Micheletti explained that higher mortality rates among enslaved men and racial whitening policies in Latin America are potential explanations for the discrepency.” (USA Today, 2020)
“Despite shared experiences one must also be careful to respect the localized context and different historical trajectories across the Afro-Diaspora. Instead of just letting one single perspective on inter-racial relationships overcloud things.” (Fonte Felipe, 2018)
The fact that Afro-descendants across the Americas as well as Cape Verdeans usually tend to display sex-biased admixture has been known for quite a while already. As several DNA studies have described it (e.g. Stefflova et al. (2011)) European DNA is usually inherited by way of male ancestors. Which is especially detectable from Y-DNA. Within 23andme’s study this topic is also extensively covered. At times with interesting outcomes (especially regional admixture on the X-chromosome, see Table S10 in the supplemental data). However the main discussion regrettably tends to be overgeneralizing by singling out only one type of explanation for their research results. Even more so lamentable because any needed nuance has usually been absent as well in the often sensationalist news reports covering 23andme’s study (such as USA Today).
Of course I am not in favour of any continued sugarcoating of all the horrible practices which took place during Slavery. And I also applaud the stated intentions of 23andme’s researchteam to “highlight the hardships people of African descent had to endure“. However at the same time I do also think that a one-sided and heavily Americanized way of framing things is not going to contribute to a balanced evaluation and complete narrative of complex histories. In particular such an approach is at risk of obscuring and distorting important localized variations across the Afro-Diaspora. Suggesting that European admixture is rape by default is potentially insulting the personal and unique family genealogies of millions of people! Including of course also many 23andme customers! One might also wonder to what extent commercial DNA testing companies such as 23andme should be involved at all in such endeavors of historically sensitive interpretation. Especially when such issues might also be politically charged. As the saying goes: never mix science with politics!
The 23andme researchteam is to be commended for admitting that they are not well-versed in the history of the Afro-Diaspora. Apparently also lacking basic knowledge of Latin American racial composition as it seems some of their frankly quite self-evident findings “surprised” them. I am not quite sure how intensively 23andme’s researchteam involved their consultants, which include several well-noted historians. But really from several clues I strongly get the feeling that a (unbiased) consultant with Latin America as main field of expertise was sorely missing. Even when given the nature of their topic a genuine diversity approach obviously was a prerequisite for describing an appropriate historical framework suitable to their research outcomes.20
Either way it is important to keep in mind that personal family histories are bound to sometimes deviate from the assumed narrative. Ultimately it might be self-defeating to allow generalizations about European admixture to determine how you should feel about your own unique DNA makeup. Especially without at least having done any basic genealogical & historical research of your own in advance. You might then of course still encounter many negative aspects. Especially when solely judging through a modern moral lens. However also often unexpected details might turn up enriching your research and making it more insightful. Things are often far more complex, inter-connected and intricate than you might assume at first. Also in the greater interest of racial healing within the USA such research efforts may be beneficial. See also:
- Coming To The Table: Genealogy Support
- Muhammad Ali: Boxer’s ancestral Irish town pays tribute after death (BBC)
- Black/White Interracial Marriage Trends, 1850–2000 (A. Gullickson, 2006) (see p.3 for an insightful chart)
4) Exciting future prospects
“We hope this paper helps people in the Americas of African descent further understand where their ancestors came from […]” Micheletti wrote.” (CNN, 2020)
“For many of us genealogists, especially working to identify our indigenous and African roots, 14-20 generations can be an extremely difficult place to reach via paper trail. This is why DNA testing is important in our search for our roots. If you took the lowest number of a generation (20 years) and multiplied that by 14 you get 280 years, minus our current year of 2021 you get back to the year 1741. For some people, they might be lucky in identifying an enslaved ancestor from that time period but potentially never identifying their origin.”(Boricuagenes, 2020)
“No records mention parents which means this is currently where the line ends for me. For many BIPOC (Black, Indigenous, People of Color) these are the issues we come across when researching and why studies like this done by 23andMe can be important in shining a light into this dark past.” (Boricuagenes, 2020)
To conclude this review of 23andme’s study I would just like to reiterate that I certainly think 23andme did an outstanding job by publishing their research findings. As far as I am concerned it is indeed helpful in all the ways quoted above. My extensive discussion of the study’s various limitations, inconsistencies and at times lack of complete/proper historical context is merely to be taken as constructive criticism. In future blog posts I aim to revisit the very valuable data-set from 23andme’s study. Which again is unprecedented especially because of its sample size! In total around 50,000 research participants. Among whom 27,422 from all across the Americas as well as Cape Verde!
DNA studies such as Micheletti et al. (2020) as well as my own surveyfindings can serve as helpful tools in your own personal quest to Trace African Roots. Although as always you do need to keep in mind inherent limitations. For those not accustomed to reading DNA studies they might seem daunting at first. Still I think that if you wish to see your personal DNA test results being placed in a wider context DNA studies can be very valuable. This is besides any general interest of course. See link below for some tips on how to “read” DNA studies.
Figure 4.1 (click to enlarge)
“Micheletti said he hopes to explore these hypotheses for the discrepancies more deeply. “We offer a lot of these potential explanations but we need to go and physically test those explanations,” he said. “We’d also like to kind of shape results and put them into a more personal context for 23andMe customers.” (USA Today, 2020)
To wrap things up I will just briefly mention this very tantalizing glimpse of what seems to be under preparation: personalized African IBD results! I do not have any further details. However if implemented as stated this could really be a game changer! I am quite excited about it. It almost seems like a reincarnation of 23andme’s former Country of Ancestry tool. But this time even specifying IBD matches with specific African ethnicities! I do also wonder how it will affect the current Recent Ancestor Location (RAL) feature. Which has frankly been underperforming when it comes to assigning African RAL’s for Afro-descendants. Mainly due to restrictions of African reference datasets as well inappropriate settings in regards to implied timeframe of ancestral connections (see section 2 of this page). I imagine both aspects will be crucial again when this new African IBD sharing tool will be realized by 23andme.21
Either way this new African IBD feature could be highly beneficial for enhancing your own narrative about your personal African roots! I sincerely hope the additional context given by 23andme to help customers understand their results will be adequate and historically accurate. And that 23andme will refrain from the temptation to only provide easy infotainment. To put it mildly their track record has not always been very positive in this regard. As always it will be important to also independently verify any claims made by 23andme. Contrasting with historical plausibility as well any other relevant factors. Remaining critical but also open-minded and careful not to be dismissive when informational value can still be obtained! Afterall it is not just about the results but also about the correct interpretation of the results 😉
Recommended viewing: Presentation of 23andme’s 2020 study (incl. FAQ):
- Presentation on Rootstech 2021: The Genetic Consequences of the Transatlantic Slave Trade, Part 1
- Presentation on Rootstech 2021: The Genetic Consequences of the Transatlantic Slave Trade, Part 2
1) The research approach of 23andme’s study from 2020 shows a great deal of overlap with my own previous and ongoing research. In spite of naturally a big difference in sample size! In 2015 I already compared 350 African American AncestryDNA results with 24 Cape Verdean ones.
“We included the Cape Verde archipelago as control region since it is an Atlantic African country that was colonized by the Portuguese in the 15th century and has been inhabited primarily by the descendants of enslaved people from Senegambia”. (Micheletti et al., 2020, p.266)
“[…] a comparison is shown between the average African breakdown for 350 African Americans and two African control groups: 24 Cape Verdeans and 7 Nigerians. Obviously the sample size of especially the Nigerians being very limited. However there are already some illuminating regional patterns to observe” (Fonte Felipe, 2015)
Because of my ongoing surveying efforts I was able to compare the same African American surveygroup (n=350) with 40 Cape Verdean samples in 2016 for my post “AncestryDNA Results Across the Diaspora“. Also including five other survey groups from across the Atlantic Afro-Diaspora actually. This was my first published blog post providing a detailed analysis of how (despite limitations) genetic data for Afro-descendants may roughly correspond with the African regions of embarkations as defined by Trans-Atlantic Slave Trade historians. In fact at that time I myself also had direct contact with the eminent historian David Eltis and other scholars behind the Trans Atlantic Slave Trade (TAST) Database. The makers of the Slave Voyages website kindly shared my 2016 research findings on Facebook. They were consulted by the 23andme research team as well for their 2020 study (see credits).
I discovered the amazing Slave Voyages website around the same time I received my own 23andme results in 2010 already. And ever since I have often relied heavily on the Trans-Atlantic Slave Trade Database as some sort of baseline. To establish historical plausibility within my ongoing research efforts on how personal DNA test results (regional admixture & African DNA matches) of Afro-Diasporans may already be in alignment with historical expectations. See these pages for an overview:
- Ancestry surveys (regional admixture: 2013-2018)
- Ancestry surveys (DNA matches: 2017-ongoing)
- 23andme surveys (2013-2020)
Before 23andme finally came through in 2018 with its current African regional set-up I had actually already performed a survey in 2013 of 23andme’s previous and much more basic African breakdown which contained 18 Cape Verdeans and 54 African Americans (see this spreadsheet). And actually my very first survey efforts date back even earlier to 2011. Based on the pioneering African Ancestry Project by Razib Khan. I shared these findings also on 23andme’s online community at that time. They can still be seen in this online spreadsheet. Therefore already in 2011 I was able to make good use of a few Cape Verdean results (two, myself incl.) as some sort of control group within a broader dataset of Afro-descendants. As afterall Cape Verdeans form a special part of the Afro-Diaspora, given that their African roots are overwhelmingly from Upper Guinea. And this should also be reflected in their DNA results (if they are any good 😉 ). In this case a tellingly pronounced genetic affiliation with “Mandenka” showed up already in 2011 (see this overview)!
In my latest survey efforts (based on both 23andme and Ancestry results) my Cape Verdean survey group has taken a more robust sample size of n=100. Quite similar to the sample size of n=121 featured in 23andme’s study. Very special therefore to see my own steadily progressing findings being corroborated and replicated by 23andme’s 2020 study. For more details see:
- 100 Cape Verdean AncestryDNA results (Fonte Felipe, 2018)
- DNA matches reported for 50 Cape Verdeans on AncestryDNA (Fonte Felipe, 2018)
- 100 Cape Verdean 23andme results (Fonte Felipe, 2021)
2) Table 1 is based on the African breakdown being scaled to 100%. Basically I performed the following calculation: % for a given African region divided by % of total African amount. In order to enable an evenhanded comparison between persons or survey groups with varying degrees of African admixture. I applied this calculation also for 23andme’s data which I collected from their supplement mmc3 (Table S7). Combining the averages for 7 USA subgroups (Gulf Coast, South Atlantic, Chesapeake, East Inland, Inland Midwest, Midwest Northern States). The overall averages I calculated for n=5785 being unweighted.
I have used the term “African Americans” in Table 1. But as far as I know the samples made available to 23andme only have self-reported their origins as being USA-born with all 4 grandparents also being USA-born within the same state. In addition the only requirement seems to have been that these profiles should have African admixture greater than 5% (Micheletti et al., 2020, p.266). Technically speaking therefore 23andme’s American (4gp) samples might not per se be self-identified African Americans. They could also be biracial, or of multi-generational West Indian, Puerto Rican or Cape Verdean background. As these migrant communities have been established in the US for a long time already, prior to WWII even. I do think 23andme’s USA 4gp samples are indeed overwhelmingly African American. But especially 23andme’s data for the Northern States and Midwest might not be fully representative for multi-generational African Americans with USA-born lineage on all sides to be traced back beyond the Civil War. As I wil discuss in greater detail in footnote 13.
3) Various important and insightful studies have been published on African American genetics. One does need to take into account some differences in methodology. Not meant as an exhaustive overview but see below for a few recommended papers. The first study by Zakharia et al. (2009) actually being quite pioneering in that it already performed a meaningful autosomal analysis of African Americans more than ten years ago. Utilizing an admittedly very limited but still well chosen African reference data-set. Including samples from both West and Central Africa. Its main findings being rather crude but still from a macro-regional perspective already quite close to my own research findings based on AncestryDNA.
- Characterizing the admixed African ancestry of African Americans (Zakharia et al., 2009)
- The genetic ancestry of African, Latino, and European Americans across the United States (Bryc et al., 2014)
- The Great Migration and African-American genomic diversity (Baharian et al., 2015)
Follow links below for a section of my blog which features DNA studies on African American genetics which I have reviewed in the past.
4) Unlike what is sometimes assumed 23andme is not just randomly dishing out results for Afro-descendants! The seemingly consistent and usually quite high “Nigerian” scores on 23andme may have seemed like a perhaps overplayed theme for especially African Americans. But actually for other parts of the Afro-Diaspora other African categories tend to show up with greater frequency. Making for a significant distinction. As shown in 23andme’s study as well as my own survey findings. See also this blogpost:
Also among African Americans there is a lot more meaningful nuance to be seen if you take into account substructure according to ultimate US state origins. More or less conforming with historical plausibility. As discussed in greater detail in section 1 of this blog post. Also to be kept in mind is that the predictive accuracy of “Nigerian” among actual Nigerians is quite impressive (around 90% for southern Nigerians, see this page).
As I have always maintained the labeling of ancestral categories is not to be taken as gospel! In all my blog posts I always use quotations to refer to categories such as “Nigerian”. And not for nothing! Because inherently there wil be some border crossing overlap. Fanning out into neighbouring areas according to some declining gradient. More insight to be gained by learning how people from various known background tend to score for these categories. Which is why I have performed my surveys among both Africans and Afro-descendants throughout the years. See also chart below for my latest findings prior to creating this page:
So for example “Nigerian” will also possibly be indicative of ancestral ties with neighbouring countries. Even when its main ancestral implication will usually indeed involve Nigeria. Furthermore while for most people southeast Nigerian (Bight of Biafra) lineage will be indicated. For some people instead it will be mostly southwest Nigerian (Bight of Benin) lineage which is being suggested. The likelihood depending on your particular Afro-Diasporan background. Either way, as demonstrated by their own study as well as my overall survey findings 23andme does not report non-sensical or randomized results. Or even just automatically assign certain regions to all Afro-descendants, regardless of their actual background!
For immediate understanding a visual depiction might be best suited. Which is why I created Map 1.1 to illustrate the wider geographic distribution of “Nigerian” (see this page for similar maps). Regrettably 23andme is still not providing such maps and other types of clarifying context which may improve the ability of their customers to make more sense of their results (unlike Ancestry). Going beyond the potentially misleading country labeling and percentages which are not always properly understood by many people.
Actually within the study’s supplemental data (mmc3 /Table S7) you can also independently verify the predictive accuracy of each of 23andme’s African regions. As they list the group averages for their African reference panel (see Table 3.1). Take note how both “Bight of Benin” samples (Yoruba) and “Bight of Biafra” samples (Igbo, Esan and western Cameroonians) show a convincing level of “Nigerian” of over 80%. However in fact also 23andme’s samples from the “Gold Coast” (incl. Ewe as well as Ashanti and Fante) still have practically 25% “Nigerian, on average! Very similar to my own West African survey findings referred above already. Which may be smaller in sample size but shows greater ethnic specification.
***(click to enlarge)
5)I firmly believe that despite inherent limitations and given correct interpretation 23andme’s regional admixture estimates can be very useful as a stepping stone for follow-up research. And just to get a general idea of where most of your African ancestors hailed from. All according to the latest state of knowledge. Which naturally may be improved upon across time. I find it important to stay positive and focus on what ever informational value you can obtain despite imperfections. Instead of taking a dismissive stance right away. Preferring to see the glass as half full rather than half empty You do need to make an effort yourself and stay engaged to gain more insight though!
In particular your follow-up research may include a focus on your African DNA matching patterns and how your African DNA matches may validate or correlate with your regional admixture scores. For example if you find around 10 African matches and 4 of them appear to be Igbo Nigerians then this solidifies and also potentially specifies any major “Nigerian” score you might have obtained. See also:
- African DNA matches reported by Ancestry for 50 African Americans (under preparation)
Furthermore you will want to expand your knowledge about the historically documented presence of Africans in your earliest known places of origin within the Americas. In order to establish the historical plausibility of your 23andme scores. For example if you happen to be African American it is vital to be aware of not only Trans-Atlantic and Intra-American Slave Trade. But also learning about Domestic Slave Trade and Post Slavery migrations will be crucial for your deeper understanding. As actually going by sheer numbers Domestic Slave Trade was most significant for the USA. An estimated 1 million enslaved African Americans (often with Virginia background) are known to have been victimized by the so-called Second Middle Passage (see this link). For more detailed discussion and references see:
- African-American Migrations (In Motion: The African-American Migration Experience)
- Slave Voyages: not only Trans-Atlantic but also Intra-American!
- The Igbo Connection for Virginia & Virginia-Descendants
Any follow-up research is of course to be customized according to your own personal situation and also according to your own research preferences. Plain genealogy is indepensable for dilligently building up a decent family tree. Which is very valuable in itself. But regrettably these strictly genealogical efforts will usually not lead you back all the way to Africa. Save for some rare exceptions (Questlove on Finding Your Roots). Hence why I always insist on avoiding any source snobbery with relation to regional admixture analysis, such as performed by 23andme.
However when duly performed your family tree research will allow you to at least identify your earliest known ancestral locations within the Americas. Which will make it easier to correlate with slave trade patterns and documented African ethnicities for those areas. And if you are very persistent and/or lucky this might also eventually allow you to find localized documentation (plantation records; private correspondence of slave owners; church records; newspaper advertisements about runaway slaves etc.) possibly even mentioning any of your African-born ancestors on 1 single family line!
Combining advanced genetic genealogy techniques such as triangulation and DNA Painter with regional admixture of shared DNA segments also holds great potential in my opinion. As it might enable you to identify an earliest family line associated with such regional admixture! Especially when this regional admixture is distinctive such an approach can be very fruitful. For example when dealing with possible Malagasy lineage the presence of any “Southern East African” and/or Southeast Asian admixture should be very useful. Naturally all of this is to be combined with any other clues you might have. Also it goes without saying that extra scrutiny is always required in order to avoid jumping to conclusions!
For some very useful blog posts with detailed instructions read:
- Pinpointing the Origin of Family’s Igbo Ancestry with DNA (Roots Revealed)
- Using DNA Painter to Verify Igbo Origins (Roots Revealed)
- Got Roots in Madagascar? (Roots Revealed)
- Genetic Genealogical Methods Used to Identify Diaspora Relatives of Members of the Kassena Ethnic Group in Northern Ghana (TAKIR: The African Kinship Reunion)
- Discovering Igbo Roots Through Genealogy and DNA (Who Is Nicka Smith)
- My Nigerian father’s DNA, African American double cousins and continental African DNA cousins (Continental and Diasporic African genetic genealogy)
- Chromosome 7 – An African American Connection (Boricua Genes)
6) According to recent research by historian Greg O’Malley (also referenced by 23andme’s study) the share of Intra-American slave trade for North America was considerable. But still overall speaking quite minor when compared with Trans-Atlantic Slave Trade (see this overview). And even more so when compared with Domestic Slave Trade. It is commendable that 23andme’s research team chose to include these recent insights about Intra-American slave trade in their study. But it is arguably also somewhat misleading to not at the same time mention the major and most likely primary genetic impact resulting from Domestic Slave Trade, mostly originating from Virginia and neighbouring states. And therefore ultimately reflecting the Trans-Atlantic Slave trade patterns for the Chesapeake to a large degree (see Figure 1.1).
I find that also in other ways the historical context of Intra-American slave trade might not always have been described in a proper way in 23andme’s study. First of all as mentioned by the makers of the praiseworthy Intra-American slave trade database their work is still very much under progress. Even when this new database is already very valuable still it will mostly not include clandestine slave voyages as well as the earliest inter-colonial slave voyages from the 1500’s/1600’s. Also “Intra-American traffic to the French Caribbean is vastly underrepresented in this database“. While “the focus on maritime voyages limits coverage of overland slave trading, a common feature of the intra-American traffic“. This goes not only for the USA, but also Latin America in fact. Which is why Figure S13. “Intra-American slave trade counts” included in the supplemental data, really should be annotated as being based on preliminary and incomplete data.
“Documented intra-American voyages indicate that the vast majority of enslaved people were transported from the British Caribbean to other parts of the Americas, presumably to maintain the slave economy as transatlantic slave trading was increasingly prohibited“. (Micheletti et al., 2020, p.272)
In regards to the quote above I am again not convinced that this is a complete or even correct representation of the historical context. Certainly not for the USA. As from my understanding Intra-American slave trade from the British Caribbean into North America was relatively most impactful during the late 1600’s and early 1700’s. A steady flow of Intra-American slave trade continued into the late 1700’s and even up till 1808 to be sure. But not because it was an alternative to Trans-Atlantic Slave Trade but rather to keep up with rising demand as well as to reach marginal US colonies not covered directly by Trans-Atlantic slave trade (see Tables 5, 8-12 in O’Malley 2014). I find it very hard to imagine that illegal slave trade after 1808 would be routed mainly by way of the British Caribbean as it is known that the British fleet carried out a strict control on such activities. Instead I believe the Hispanic Caribbean (esp. Cuba) may have been intended (see also this paper).
Furthermore I think it is a bit too simplistic to only single out Nigerian captives as being most prominent within Intra-American Slave Trade. Especially for South Carolina this additional flow of African captives arriving by way of the West Indies will certainly have diversified the regional origins of its enslaved population. Increasing its share of Nigerian ancestry. Beyond what might have been expected when only looking at Trans-Atlantic Slave Trade records. However in addition to the Bight of Biafra also captives from the Bight of Benin (usually Gbe speakers from Benin and Togo up till the late 1700’s) and the Gold Coast are known to have predominated slave trade from the British Caribbean. In particular in the early phase of colonization (late 1600’s/ early 1700’s). See also chart below taken from O’Malley 2014.
In previous blog posts I have speculated for example to what extent “Benin/Togo” scores reported by Ancestry for African Americans might (partially) be explained by way of Intra-American Slave Trade and the arrival of Gbe speaking people from modernday Benin, Togo and eastern Ghana. Or (probably less common) also by way of illegal slave trade occurring after 1808. Originating from the Bight of Benin and at times also including Yoruba captives. Their genetic inheritance will mostly be translated into “Nigerian” scores on 23andme I suppose. Given the absence of an equivalent of the “Benin/Togo” category within 23andme’s African breakdown. For more discussion:
- How to account for unexpectedly high “Benin/Togo” scores among African Americans & Anglo-Caribbeans? (Fonte Felipe, 2018)
7) For those who cannot access Dr. Fatimah Jackson’s article in the American Journal of Human Genetics I will provide a short summary of its main arguments. I will also briefly discuss the response by 23andme’s researchteam. Dr. Fatimah Jackson’s critique is also mentioned in this freely available article.
- Artificiality of the colonial map in Africa
- Regional variation in the proportions of Africans brought from different regions into the staging areas of North America.
- Domestic Slave Trade introduces distortions into the African ancestries of African North Americans
- The role of rebellions in distorting African survivorships patterns.
- The immediate post emancipation period created opportunities for the emergence of African North American microethnic groups.
- Post enslavement migrations were significant initiators of displacement in African North Americans.
- Yoruba bias in public genetic databases.
Especially arguments 1-3 are greatly in line with my own reasoning as featured in various blogposts since atleast 2015. In response 23andme’s research team stated the following:
“On one hand, literature discussed by Jackson, such as displacement of populations via inter-colonial migrations and the Great Migrations, are addressed in Micheletti et al. On the other, literature describing the tendency for certain African populations to rebel more than others, or the biased forced breeding of certain enslaved populations, are not included in Micheletti et al., but provide additional historical context consistent with their observations.” (Micheletti et al., 2021, p. 209)
Basically conceding that 23andme overlooked the additional (and most likely greater) role played by Domestic Slave Trade in shaping the African American genepool and its main regional African tendencies. To their credit they also mentioned this during a presentation made for Rootstech 2021 (see this video). As far as I know the approximate extent of intentional breeding practices in Virginia and neighbouring states is still to be determined. However it is certainly needed to also explicitly mention this horrifying aspect of Domestic Slave Trade. Aside from plentiful documented evidence some African Americans might still have some lingering family lore about this gruesome and dehumanizing practice. For example:
- McGruder roots: Families trace lineages back to a single enslaved man (ABC News, 2021)
- My Slave Sister Myself (documentary, 2009)
While contemplating the many wider implications of “slave breeding” it might be advisable however to keep in mind that Virginia was also the birth place of many magnificent facets of currentday African American culture. Furthermore the dispersal of free Afro-Virginians into the Midwest and the Northeast, both before and after Slavery, is worthy of further exploration as well. For a quick introduction:
- African Nation Founders: Africans in the Chesapeake (NPS – Ethnography Program)
- African American Ties to the Chesapeake Bay Region’s History, Culture, and Waterways (Saving Places, 2021)
- African American Resources for Virginia (FamilySearch)
In their reply to Dr. Fatimah Jackson 23andme’s research team reject the suggestion that their study design would be fundamentally flawed. Something I agree with for the most part. In particular with regards to the African data-set being used in this study I would say it certainly seems to be carefully chosen and quite representative of Atlantic Africa. Although probably imbalanced by a greater and more varied selection of Central African samples (see Table 3.1). Of course improvements are always possible (addition of Southeast African samples as well as from West Africa’s interior, Benin/Togo etc.). But at times more difficult to achieve than imagined. Therefore I believe that credit should also be given whenever there is a genuine push towards greater coverage. Which is certainly true with this study by 23andme which also includes a highly relevant number of Angolan samples!
8) Actually other earlier studies, based on autosomal analysis, have also already been suggestive of a lower than expected genetic contribution from Senegambia among African Americans. See for example:
- Characterizing the admixed African ancestry of African Americans (Zakharia et al., 2009)
- Genome-wide Comparison of African-Ancestry Populations from CARe and Other Cohorts Reveals Signals of Natural Selection (Bhatia et al., 2011) (see also this figure)
- Unravelling the hidden ancestry of American admixed populations (Montinaro et al., 2015)
9) Relying on either estimates or documented slave voyages the share of Senegambia in direct Trans-Atlantic Slave Trade into North America would be around 20-25%. Also depending on whether to include or exclude slave voyages with unknown provenance (“Other Africa” represents a share of about 20%-25% in the Trans-Atlantic Slave Trade Database, see also this screenshot). See section below for more similar charts:
***(click to enlarge)
Interestingly when 23andme rolled out their new African breakdown in 2018 their website mentioned that “Today, around 20% of African American ancestry is from this region“. Referring to “Senegambian & Guinean”. Undoubtedly based solely on Trans-Atlantic Slave Trade patterns. An illustration of widespread overestimation of Senegambian ancestry among both American academics and the general public. Understandable given the dominant historical narrative sofar. However as I already suggested in 2016, based on my AncestryDNA survey, this conventional idea is now up for reconsideration.
***(click to enlarge)
10) This topic of lower Senegambian lineage than expected among African Americans is to be viewed from various angles. As I suggested already in 2015 it is also to be kept in mind that quite likely “Upper Guinean origins beyond Senegambia in a strict sense, either from the deep interior or from areas to the south, are relatively more important for African Americans.” Which is to say that lineage from Sierra Leone and western Mali might be more pronounced than strictly Senegambian lineage. On AncestryDNA (2013-2018 version) this seems to have been implied by a greater frequency of primary “Mali” scores, as well as higher outliers for “Mali”, when compared with the “Senegal” region (see this chart). For more discussion on how to zoom into the specifics of your Upper Guinean DNA see:
Within my own African American 23andme survey (n=200) the primary scores (1/200) as well as maximum scores (20.4%) for “Senegambian & Guinean” were again indicative of a strong degree of dilution. Regrettably I could not find any such data within 23andme’s study. In 23andme’s current African breakdown there is also no direct equivalent for Ancestry’s “Mali” region. Which is why I suspect that additional DNA from Mali and surrounding countries will be covered by “Ghanaian, Liberian & Sierra Leonean” as well (see this map and also this page for the 23andme results of a few (mixed) Malian persons).
Unfortunately it seems that after 23andme’s latest update in 2020 “Senegambian & Guinean” scores may have become underreported for some people. Due to 23andme’s upgraded “box car” algorithm being biased towards diluted admixture segments. 23andme is still able to detect and pinpoint distinctive types of minor regional admixture. But it will now tend to assign them to related but more prevailing admixture categories whenever the segments happen to be adjacent and located on the same chromosome. In particular a submerging of actual Senegambian segments into “Ghanaian, Liberian & Sierra Leonean” scores might occur I suppose. See also:
11) Obviously my preliminary survey findings were often based on minimal sample size and also other shortcomings might apply. However for the most part the underlying trends I identified in 2015 already have now been vindicated by 23andme’s much larger and extensive study from 2020, as illustrated by Table 1.4. For more discussion of my previous substructure findings among African Americans see:
- Regional variation according to USA state origins? (scroll down to section 5) (AncestryDNA 2013-2018 version)
- Substructure according to state origins (23andme 2018 version)
- South Carolina 23andme results (23andme 2020 version)
Also interesting to compare this sheet below featuring the pioneering 2013 findings of Ancestry’s research team (see this presentation) with Map 1.3. taken from 23andme’s 2020 study. In both cases South Carolina appears to have the greatest frequency of high outliers for Senegambian DNA. Something which I have observed as well in both my Ancestry and 23andme surveys. Although additionally also Louisiana Creoles at times are able to obtain clearly above average scores for “Senegambian & Guinean”. Although generally speaking (due to both Intra-American and Domestic Slave Trade) the regional African origins of people from Louisiana are often quite similar to African Americans from other parts of the USA.
***(click to enlarge)
12) Based on Table 1.4. one can indeed pick up both on some nuanced degree of differentiation on the one hand. Suggestive of historically plausible substructure in especially South Carolina. But on the other hand 23andme’s data-set is still also featuring a relative consistency of primary “Nigerian” scores among African Americans across the country. While also otherwise the regional proportions are quite similar, at the very least rankingwise.
I find it interesting to contrast this with my observations from a few years back, quoted below. Based on more variable AncestryDNA results (confusingly also between siblings and close family!) Of course not entirely comparable because of a more detailed African breakdown on Ancestry as well as a less “smoothing” algorithm at that time (2013-2018). At the risk of oversimplifying things I do think that the seemingly more consistent regional proportions on 23andme are more in line with a largely recognizable genetic profile for African Americans. And also with its main historical evolution. Despite inevitable individual variation and at times distinctive localized substructure.
“you might expect somewhat steadier proportions of each African region being reported in each single result. Instead what seems to be showing up is a great deal of heterogeneity despite African Americans arguably forming an interrelated ethnic group of their own for several centuries. With a relatively small founding population compared with other Afro-diasporic groups because of natural increase already occurring in the mid-1700’s. And additional African geneflow being relatively limited especially after Abolition of the Slave Trade in 1808″ (Fonte Felipe, 2015)
“When I first started out my survey (already 4 years ago) I was under the assumption that African Americans would be much more similar in their regional components. I was expecting a blended but still rather consistent African breakdown on average. Due to so many generations of Africans Americans residing and intermarrying within the USA (atleast since the early 1700’s/late 1600’s) and also being derived of a relatively small African founding population (in comparison with other parts of the Afro-Diaspora).” (Fonte Felipe, 2017)
13) As already pointed out under footnote 2 it seems that 23andme’s American (4gp) samples might not per se be self-identified African Americans. But rather either biracials or people with multigenerational migrant origins. Afro-descended indeed (atleast African > 5%) but without USA-born lineage on all sides to be traced back beyond the Civil War. Quite unfortunate as I suppose with adequate quality control this kind of “data contaminination” could have been prevented. But possibly arising from confusion on how to define African Americans an an ethnic group versus “Black Americans” or just Afro-descended Americans in an even broader sense. See for example:
- African DNA and African American Identity (23andme)
- Facts About the U.S. Black Population (10% foreign-born; 8% multiracial non-Hispanic; 5% Black Hispanic) (Pew Research Center, 2021)
I greatly suspect that this might especially be an issue for 23andme’s data for the Northern States and Midwest (which includes Texas). This can be most easily deduced from their atypical levels of non-African admixture. Which can be verified by consulting the study’s supplemental data (mmc3 /Table S7). Overview below taken from that supplement. Everything highlighted in red seemingly suggestive of either Hispanic American or even Cape Verdean-American samples being included to a great (and distorting) degree. Even impacting the level of “Senegambian & Guinean” in fact. Which (when scaled) is above average for the Midwest and esp. the Northern States when compared with other parts of the USA.
Another part of 23andme’s supplemental data (Table S8/mmc4) seems to suggest that as much as 7% of their USA samples are not likely to be African Americans. As indicated by atypical continental haplogroup combinations and even more so the associated admixture levels. For example about 4% of 23andme’s USA samples have both paternal and maternal European haplogroups. Which is not entirely impossible in individual cases (e.g. Henry Louis Gates Jr). However the average degree of African admixture for this group of people is only around 26%! I suppose for future studies preventing this sort of sampling errors should be part of the standard procedure. Especially when aiming to elucidate ancestral narratives for African Americans. Given the general demographic tendencies in the USA also a more thoughtful stance should be taken then in regards to the ethnic affiliation of African Americans with USA-born lineage on all sides to be traced back beyond the Civil War.
***(click to enlarge)
14) “More is better” is a very current belief. Not only in DNA testing but also generally speaking. However this assumption does not always hold true. I would argue that quality of content should come first. Unlike commonly assumed you do not need to sample entire populations to obtain informational value with wider implications. Naturally greater sample size does (usually) help matters. But if you randomly test a given population, and if your sample group is fairly representative of the whole population, you can already make meaningful inferences. Naturally methodology applied and the assumptions being made should be made explicit, but this is common scientific practice. See also:
- Representative Samples: Does Sample Size Really Matter? (SurveyGizmo)
This is an important lesson I learnt while performing my previous AncestryDNA surveys: robust patterns (in line with historical plausibility) might already be discernible from a sample-size of around n=30. Which is actually often considered a general rule of thumb. Adding more results will indeed lead to greater finesse and more detailed statistics but the main outline might then already be established. Even more so when you are aware of any possible sampling bias or substructure and know how to account for it in your analysis. And in fact officially published studies based on much larger sample size have usually vindicated or confirmed my own findings. While due to free format on my blog I am often able to provide greater detail and more appropriate context.
This goes for example for my previous survey of 350 African American AncestryDNA results which I published in 2015 already. I was able then to independently verify the coherency of my regional ranking patterns for not only my overall African American survey group. But in fact also for my Louisiana, Virginia and South Carolina survey subgroups! Because at that time Ancestry publicized the summarized Ethnic Estimates for over 250.000 of their American customers! By way of Ancestry’s so-called “Genetic Census of America” I was able therefore to find confirmation by way of Big Data! Other survey findings of mine on both Ancestry and 23andme have also been independently confirmed or I suspect will be confirmed eventually in upcoming academic studies.
15) In my previous AncestryDNA surveys for Africans & Afro-Diasporans (2013-2018) I have also made good use of a macro-regional format (see this chart). Which is still sub-continental. Basically combining interrelated & neighbouring genetic regions in order to allow certain regional patterns to show up more clearly. This turned out to be particularly helpful when wanting to explore any rough correlation between aggregated AncestryDNA results and slave trade patterns. Which I indeed established for the most part.
In my 23andme surveys I have again adopted a macro-regional framework. Combining overlapping regional scores from within West Africa versus Central & Southern Africa versus Northeast Africa (see this overview). More basic than the actual African breakdown on 23andme (featuring atleast 12 categories). But I find such an approach to be potentially quite insightful as it enables an intermediate perspective which is often easier to make sense of. I found it especially helpful to realize that Central African (related) DNA goes beyond most likely underestimated “Angolan & Congolese” scores.
However due to fewer West African regions available on 23andme this macro-regional set-up is not completely the same as what I have used for my previous Ancestry surveys (Upper Guinea, Lower Guinea, Central & Southeast Africa, see this map). Similar regions to “Mali” and “Benin/Togo” are lacking on 23andme at this moment. Another big difference is caused by 23andme’s more conservative approach than on Ancestry. Whereby African DNA which cannot be classified reliably (given 23andme’s limited set-up) is put under either “Broadly West African”, “Broadly Congolese & Southern East African”, or “Broadly African”. Such scores are often exceeding the double-digit level even (when scaled)!
Follow link below for better understanding of how a macro-regional perspective can be beneficial when looking into DNA results. Because it takes into account the various intersections based on historical and ethno-linguistic considerations, aside from merely genetic ones. Mutually reinforcing but still only meant to be indicative of course and to be used as proxies!
- African macro-regions, based on slave trade history, ethnolinguistics and genetics (using Ancestry’s regional format)
16) Within 23andme’s study also European and Native American admixture is being analyzed. However other types of continental admixture are being left out of scope. In my own survey findings I have found it useful to also include statistics for other categories, such as North African, as well as West Asian, South Asian, East Asian and Southeast Asian. When compared with also other published studies my inclusion of Southeast Asian admixture statistics might actually be a first, as far as I know. In fact I also analyzed likewise minimal yet distinctive levels of South Asian (0.4%) and Native American admixture (0.3%) among Cape Verdeans. Interestingly the latter finding is also appearing in 23andme’s study, albeit even more so as a trace amount (0.1%). See also these overviews:
- Continental breakdown on 23andme according to my own survey findings (Fonte Felipe, 2021)
- Continental breakdown according to 23andme’s study (Micheletti et al., 2020, supplemental data Table S7)
17) For a greater understanding of this Upper Guinean Founding Effect read the following blog posts:
- Shared Upper Guinean roots between Cape Verdeans and Latin Americans (Fonte Felipe, 2014)
- Dominican AncestryDNA results (Fonte Felipe, 2015)
- Afro-Diaspora AncestryDNA results: A Comparison (Fonte Felipe, 2016)
- Update: Afro-Diasporan AncestryDNA Survey (part 1) (Fonte Felipe, 2018)
- Fula, Wolof or Temne? Upper Guinean AncestryDNA results 2013-2018 (Fonte Felipe, 2019)
An overly USA-centric perspective may have prevented a full realization of how significant Upper Guinean ancestry turns out to be for many Hispanic Americans. Especially in comparison with African Americans. The recent inclusion of early Iberian (Portuguese/Spanish) Slave Voyages into the standard reference Slave Voyages database has been incredibly useful therefore for greater understanding. However it should be pointed out that Latin American (e.g. Carlos Esteban Deive, Gonzalo Aguirre Beltrán), Iberian and Cape Verdean historians (such as António Carreira) have always been aware of the significance of this early slave trade by way of Cape Verde. Their research findings may not have been so widely known in the USA merely because their work has mostly not been published in English.
18) It is a bit peculiar that within 23andme’s study only the “underrepresented” Senegambian IBD findings for Hispanic Americans are discussed with more detail. I mainly agree with the reasoning (greater dilution and older segments because of Upper Guinean ancestry generally to be traced back to the 1500’s/1600’s). But personally I would say 23andme’s regional admixture findings are much more solid and convincing. With a further expansion of 23andme’s African Reference data-set (including more ethnically diverse Upper Guinean samples) as well as further tweaking of other aspects I suspect 23andme’s IBD findings are likely to change quite a bit. Either way their main finding of Senegambian DNA segments often being older and smaller-sized for Hispanic Americans was already reported in 2013. See also:
- Reconstructing the Population Genetic History of the Caribbean (Moreno-Estrada et al., 2013)
Also in light of the recent upgrade of 23andme’s algorithm it might be more productive to take into account what kind of effect it will have if “Senegambian & Guinean” DNA segments are indeed likely to be smaller-sized and fragmented. Afterall 23andme’s upgraded “box car” algorithm is said to be biased towards diluted admixture segments. To be sure 23andme is still able to detect and pinpoint distinctive types of minor regional admixture. But it will now tend to assign them to related but more prevailing admixture categories whenever the segments happen to be adjacent and located on the same chromosome. Which is why “Senegambian & Guinean” scores may have become underreported for some people after the 2020 update. Something similar actually also took place on Ancestry after its 2019 update. For more details:
- Oversmoothing algorithm? (scroll down to section 2) (Fonte Felipe, 2019)
- Update of 23andme’s African breakdown (2020)
19) I do suspect that the sampling imbalance in 23andme’s African reference dataset (see Table 3.1) has influenced its IBD findings. I described something similar when Ancestry’s African Reference Panel was updated in 2018. See:
- More is not always better: over-sampling for “Cameroon, Congo & Southern Bantu”, “Benin/Togo” and “Mali” causing inflated scores? (Fonte Felipe, 2018)
But I do still find it commendable that an effort was taken to include historically relevant African samples. I also appreciate how the ethnic backgrounds of these samples were specified. Of course improvements are always possible. But at times more difficult to achieve than imagined. Therefore I believe that credit should also be given whenever there is a genuine push towards greater coverage. Which is certainly true with this study by 23andme which also includes a highly relevant number of Angolan samples!
Complete coverage of Africa’s ethnic diversity is preferred of course but I do not believe that would be a strict precondition for producing useful research results (either IBD or admixture). Instead it is the internal coherency of the reference samples as well as the overall sampling strategy to select the proper reference populations (in balanced proportions) which make the difference! When thinking about improvement of 23andme’s African reference dataset the following comes to mind:
- addition of samples from West Africa’s interior (Mali, Burkina Faso, Gur speakers)
- addition of Gbe speaking samples from Benin
- addition of Southeast African samples, incl. Madagascar and preferably Makua samples from Mozambique!
- greater variation of Senegambian samples (whenever possible to be obtained from MalariaGEN database)
20) Quotations below are illustrative of the lack of historical nuance within 23andme’s study. Understandable perhaps given that 23andme’s researchteam is composed of geneticists who at times are also performing biological research. No matter how well-intentioned it does beg the question to what extent historically and even politically sensitive narratives should be dealt with by commercial DNA testing companies.
“Regional differences may be due to higher mortality in enslaved males in Latin America as well as a common practice called branqueamento, or racial whitening, which involved women marrying lighter-skinned men with the intention of producing lighterskinned children. National branqueamento policies were implemented in multiple Latin American countries, funding and subsidizing European immigrant travels with the intention to dilute African Ancestry through reproduction with light-skinned Europeans” (Micheletti et al., 2020, p.273)
“While the team relied on two historians well-versed in the transatlantic slave trade—Linda Heywood and John Thornton of Boston University—to help identify which parts of Africa needed to be included and to interpret the study’s findings, Jackson says this study suffers from a common issue: while “the authors are to be commended” for their contributions, geneticists often fail to include historians early in the planning of their studies, bringing them in only at the end to explain anomalies. “Geneticists, like any scientist, believe if we have high quality data, the data will speak to us,” Jackson tells The Scientist. Instead, what is needed is for “geneticists to work more intimately with scholars who have knowledge of the historical and social context of evolutionary processes.” (The Scientist, 2020)
To be sure intentional Europeanization policies by way of migration did exist in a few Latin American countries. Mostly implemented however in the 1800’s after having gained independence. Argentina and Brazil possibly being the best known examples. As far as I know the effect on any increase of actual racial miscegenation requires further research though. As actually most of these European migrants (esp. when not Portuguese) remained relatively endogamous. Residing mostly in their own communities set up in southern Brazil. Leaving northeast Brazil with its huge but also heavily admixed Afro-descended population mostly unaffected. Without proper contextualization such statements are therefore clearly at risk of overgeneralizing things. In particular not accounting for widespread racial mixing, a.k.a. mestizaje taking place already during the 1500’s-1700’s!
Not in only in Brazil but all across Latin America actually. Places like the Dominican Republic are known to have had substantial mixed-race (“Mulato”) population segments already in the 1600’s. Solely mentioning European migration policies as a factor behind racial admixture in Latin America is obviously a grave simplification of a much greater and complex story of Latin American ethnogenesis. It actually stands in the way of deeper understanding of African regional roots among Latin Americans. Because substructure according to degree of African admixture might be quite widespread across Latin America, judging from my survey findings. And therefore the question of when approximately most of the African geneflow took place (1500’s/1600’s vs 1700’s/1800’s) becomes quite essential. For more details:
- European immigration to Brazil (Wikipedia)
- The Population of Brazil in the Late Eighteenth Century: A Preliminary Study (D. Alden, 1963)
- Freedom in the Colony of Santo Domingo (Dominican Roots, 2020)
21) Just to name one example of misleading infotainment when so much more helpful context could be given in regards to haplogroups:
- You share an ancient paternal lineage with Pharaoh Ramesses III (male haplogroup report by 23andme)
I really do want to give 23andme the benefit of the doubt. But I fear that 23andme’s researchteam might often be overruled by more pressing commercial considerations from higher management. See also:
- New 23andMe Experience – In a Word, Disappointing (DNA-Explained, 2016)
- Misleading 23andMe Paternal Haplogroup Emails For Females (DNA-Explained, 2021)
To end on a hopeful note I am really excited about the prospect of personalized results featuring African IBD. Especially given that I have always been a huge fan of 23andme’s former Country of Ancestry Tool. I would be especially thrilled if the various helpful controls to set IBD thresholds etc. will be reinstated. Also the display showing the shared DNA segments per chromosome allowed for a great deal of flexibility in exploring your DNA matches!
“Despite all of the unknown variables involved I have always felt that with careful interpretation you can still obtain a great deal of informational value from 23andme’s CoA tool. I myself was able to connect with my first Cape Verdean DNA cousin through my CoA results (way back in 2010 already!). Also several Afro-Diasporans were able to find their very first African DNA matches by way of CoA. Going beyond individual results I have always been fascinated by the potential this CoA tool could have to uncover […] generalized matching patterns between Africans and Afro-Diasporans. Being very anxious to learn if these patterns roughly corroborate what we know from historical sources and cultural retention.” (Fonte Felipe, 2018)