10/01/2026
~ Critique of Dr. Kurin's 2026 Smithsonian DNA Study regarding the Jefferson-Hemings question ~
Dr Kurin's Jefferson Book & Preprint
By John Hamilton Works Jr
30 September 2026
On 22 September 2026 the Smithsonian launched Dr Richard Kurin's book History by a Hair, which reports that DNA has resolved a 200 year old paternity question--whether Thomas Jefferson fathered children with Sally Hemings. The night before, the underlying scientific paper appeared as a preprint, The DNA legacy of Thomas Jefferson (bioRxiv, 21 September 2026). A preprint is a draft posted publicly before peer review. That paper is still under review and has not been published.
I have read both in detail. This assessment draws on Chapter 20 of the book, where Dr Kurin sets out the DNA evidence, and on the preprint itself; where the 2 report different figures, I use the preprint's.
What Dr Kurin's team accomplished is real. They appear to have recovered a usable genome of Thomas Jefferson from 200 year old rootless hair, which few laboratories in the world could have done, and they compared it directly to living descendants of Thomas Jefferson and Sally Hemings. That is the comparison this question has always needed, and they deserve credit for making it.
But the proof built on that genome does not yet meet the standard genealogists apply, and the public is being told the question is resolved when the study's own numbers do not say so. Here is what I believe people should know.
~ Disclosure of interest:
I am a 7th generation acknowledged lineal descendant of Thomas Jefferson, a member of The Monticello Association and a past president, the petitioner in a disinterment proceeding now pending in Nashville, and the principal of a separate DNA study of this question conducted with the Max Planck Institute for Evolutionary Anthropology (the Max Planck Institute). I have an interest in this topic, and readers should weigh what follows with that in view.
~ The standard the study adopted:
The preprint frames its own purpose in the terms I set in 2002, 24 years ago, and cites me for them. Reference 12 is my Discussion of the Lack of Documentary Evidence of Paternity of the Hemings Family (The Monticello Association, 2002), and the text states that the standard was "clear and convincing evidence" and that the only way of meeting it was to obtain Thomas Jefferson's DNA and compare it directly to Hemings descendants. That was the right standard then and it still is now, and they have adopted it. The question now is whether their study design and its ex*****on actually meet it.
~ The publicity came before the science:
A book, a Smithsonian event, and a magazine feature reached the public first. The scientific paper, which is under review at the Proceedings of the National Academy of Sciences (PNAS), has not yet passed peer review. The genome itself--the single most important piece of evidence--has still not been deposited in any public repository where other scientists can download it, although Dr Kurin said publicly that it would be; the preprint offers it only on request. Until it is deposited, no one outside Dr Kurin's study team can check any of this. That is not a procedural quibble; instead it is the difference between a finding and an announcement. It also reverses the ordinary sequence, in which a paper passes peer review--by reviewers chosen by a journal's editor, not by the authors--before its conclusions are presented to the public. Peer review is also no guarantee of a headline that holds up. Foster's 1998 study was rejected by Science before Nature published it, and within 2 months the authors wrote to that journal to say that "the title assigned to our study was misleading in that it represented only the simplest explanation of our molecular findings", and that their data could not exclude Randolph Jefferson's family (Foster and colleagues, "Reply: The Thomas Jefferson paternity case", Nature 397:32, 7 January 1999).
~ Monticello has not adopted the study's certainty:
Shortly before the book launch, the Thomas Jefferson Foundation, which owns and operates Monticello, published a statement by its President, Dr Jane Kamensky, and its Vice President for Research, Dr Andrew Davenport (monticello.org, September 2026). It does not call Dr Kurin's study proof. Its only endorsement of the new evidence is conditional--"Should peer-reviewed scientific publications verify the results of Dr. Kurin's new study, the new genomic evidence will further reinforce the conclusion supported by these decades of scholarship". It repeats none of the 99.9% language. For the paternity conclusion itself it relies on the Foundation's own scholarship since 2000, citing among other evidence Israel Gillette Jefferson, born into slavery at Monticello, who said in 1873 that he could "as conscientiously confirm" Thomas Jefferson's paternity as he could "any other fact which I believe from circumstances but do not positively know". The institution with the most invested in that conclusion has, in effect, said it will wait for peer review.
~ The headline number is not the number for Eston Hemings:
The figure being quoted--odds of less than one in a thousand--is the probability that Thomas Jefferson fathered at least one of two of Sally Hemings' sons, Madison and Eston. But it is not the figure for either son on his own. For Eston Hemings--the child the famous 1998 DNA study was about (Foster and colleagues, "Jefferson fathered slave's last child", Nature 396:27–28, 5 November 1998)--the paper's own Table 1 gives Thomas Jefferson an 82.6% to 86.5% probability of being his father, and leaves his brother Randolph Jefferson a 12% to 16% probability. The authors call that "substantial" or "positive" evidence, and they do not call it proof. For Madison Hemings the same table gives Thomas Jefferson 99.76% to 99.80%, and Randolph Jefferson 0.24% to 0.20%. That is why the Madison Hemings line carries the headline--and why the long shared segments behind it matter so much. For comparison, Virginia requires a scientifically reliable genetic test affirming "at least a ninety-eight percent probability of paternity" before the test establishes paternity (Va. Code § 20-49.1(B)(1)), and many states set a similar threshold. The statute governs living parentage disputes rather than historical attribution, so it binds no one here. But it is the standard an American court applies when the question is whether a particular man fathered a particular child, and the study's own figure for Eston Hemings does not reach it.
~ The paper's summary goes further than its own results:
Its introduction reports "strong evidence that Thomas Jefferson himself, not a close relative, fathered multiple children with Sally Hemings". Its results section says the father "must have been a close relative of Thomas Jefferson if not Thomas Jefferson himself", and that for Eston Hemings the Randolph scenario "retains a smaller but considerable proportion of the posterior probability". Those are not the same claims, and the press has largely reported the first claim. The summary of the 3 reviewers' comments--all hand-picked by Dr Kurin and not independently selected by the editor of a peer reviewed article--states that the likelihood ratios put paternity of Madison and Eston Hemings at over 99%, but the study's own figures do not support that for Eston Hemings. The 99.9% figure is the aggregate probability that Thomas Jefferson fathered at least one of the two sons, which is a different claim from either individual one, and it is the figure most often repeated in the press.
~ The strongest claim depends on a method used in Dr Kurin's study for the first time:
To find shared DNA in a genome this degraded, the team wrote its own computer method instead of using the published, tested one scientists rely on--ancIBD, developed at the Max Planck Institute and published after peer review (Ringbauer and colleagues, "Accurate detection of identity-by-descent segments in human ancient DNA", Nature Genetics 56:143–151, 2024). The preprint itself cites that paper. That is a departure in design, not merely a step left unvalidated, and it is used here for the first time on a historical figure of the importance of Thomas Jefferson--not tried first on pilot paternity cases of lesser prominence, as would customarily be done. The authors did check the method against the known descendants in their own panel, but a check inside the same study is not independent validation.
The strength of the Madison Hemings result depends on a small number of long shared stretches of DNA, because the paper's statistical model weighs not only how much DNA is shared but the length of each shared segment. And the preprint itself reports that where both methods were run, its own method was "generally spanning gaps between ancIBD segments"--in plain terms, where the validated method sees several shorter pieces, the team's method joins them into one long one. The book reports one Madison Hemings descendant sharing a single unbroken segment of 117 cM with Thomas Jefferson, which is exceptionally long for a relative 5 generations removed. Whether segments like that are real, or are shorter pieces joined together by a new method, is the question on which the strongest claim in the paper depends--and no one can settle it without the data.
~ Short segments count toward the result:
Dr Kurin's study counts shared segments as short as 5 cM, and reports its results at both a 5 cM and a 10 cM threshold. Short segments are the most prone to false matches. John Walden assembled 9,000 autosomal kits and tested whether segments matching a child also matched a parent--with both sides phased, 14% of 5 cM segments proved real, 42% at 7 cM, 86% at 10 cM and 97% at 12 cM ("Identical by descent", ISOGG wiki, reporting John Walden's results). At Dr Kurin's lower threshold, roughly 6 of every 7 segments would be expected to be false matches rather than evidence of a shared ancestor. Those figures come from phased modern genotypes rather than from a 2-fold genome recovered from 200 year old hair, where there is no reason to expect better performance and reason to expect worse. The Monticello Association itself sets a 12 cM minimum when deciding whether an applicant descends from Thomas Jefferson (Membership Protocol, section 3.1.3, adopted 2024); I drafted that protocol with Phil Goff, whom the Association engaged to work with me on it. But the longer segments matter more here, and it is their length that is in question.
~ The genetic yardstick was European:
To judge whether a stretch of matching DNA reflects descent or mere chance, the method needs to know how common each genetic variant is in the population. The study used a panel of people from England and Scotland, and reports similar results with a second panel of Northern and Western European ancestry. The authors explain that these frequencies stand in for Thomas Jefferson's own genetic background, and they checked for false matches against unrelated people of African American ancestry. That is a reasonable answer. What the paper does not report is whether its results change when an admixed reference panel is used--which matters because the Hemings descendants are of mixed African and European ancestry. There is a related and well-known issue in ancient DNA work generally. A degraded genome is read by lining its fragments up against a modern reference genome, and that process tends to pull the result toward the reference. It is not a flaw peculiar to this study, but it has to be measured, and the preprint's main text does not address it.
~ How easily a reconstruction can be presented as the thing itself has recently been in the news:
In April 2025 Colossal Biosciences, where Dr Beth Shapiro--this study's corresponding author--is chief science officer, announced that it had brought back the extinct dire wolf (the underlying genome study is Gedman and colleagues, "On the ancestry and evolution of the extinct dire wolf", bioRxiv, April 2025, doi:10.1101/2025.04.09.647074, since published in Cell Genomics). The animals proved to be gray wolves carrying 20 gene edits. Dr Shapiro herself later said that "our animals are grey wolves with 20 edits that are cloned" (interview with New Scientist, reported in Live Science, 23 May 2025), and the Canid Specialist Group of the International Union for Conservation of Nature's Species Survival Commission concluded that "the three animals produced by Colossal are not dire wolves" ("Conservation perspectives on gene editing in wild canids", 18 April 2025). I raise this not to impugn the science of her Jefferson work, which I have not seen, but because the same laboratory has recently been criticized for presenting a reconstruction as the thing itself--and the Jefferson genome is itself a reconstruction, dependent on the modern references and panels used to build it.
~ The other candidates were never tested--Randolph Jefferson:
Thomas Jefferson's brother Randolph, a widower living 20 miles away throughout the relevant years, was never sequenced. He was no distant figure at Monticello. Isaac Granger Jefferson, who had been enslaved there, remembered that Randolph "used to come out among black people, play the fiddle & dance half the night" ("Life of Isaac Jefferson of Petersburg, Virginia, Blacksmith", as recorded by Charles Campbell in 1847, published in Rayford W. Logan, ed., Memoirs of a Monticello Slave, University of Virginia Press, 1951). For Eston Hemings in particular, Randolph Jefferson fits the record. His first wife had died, probably about 1799, and he did not remarry until about 1809--a span covering the births of Harriet, Madison and Eston Hemings. On 12 August 1807, within the likely window of Eston's conception (Eston was born on 21 May 1808), Thomas Jefferson wrote inviting him to Monticello: "Our sister Marks arrived here last night and we shall be happy to see you also" (Thomas Jefferson to Randolph Jefferson, 12 August 1807, University of Virginia Library). Whether he came is not recorded. He was still fathering children late in life--his son by his second marriage was conceived shortly before his death in 1815, at the age of 59. In August 1807 Thomas Jefferson was 64, and that March he had been confined for about 2 weeks by one of his recurring headaches ("I am now in the 8th day of periodical head-ach which threatens to be obstinate", Thomas Jefferson to John Wayles Eppes, 21 March 1807).
~ Randolph Jefferson is separated from Thomas Jefferson only by computer simulation, and a brother shares roughly half the same DNA:
The point on which the whole conclusion depends is therefore a model standing in for the comparison sample the study lacks. That follows the design the study set for itself in 2015, when its consent form described its purpose as determining "if the former president fathered one or more of Sally Hemings' children"--a yes or no question about one man, rather than a comparison among the candidates. A bona fide paternity study keeps an open mind about who the father is, and tests every likely candidate. The Genealogical Proof Standard requires exactly that--reasonably exhaustive research, and the resolution of any conflicting evidence, before a conclusion is reached (Board for Certification of Genealogists, Genealogy Standards, 2nd edition). The international forensic standard is to the same effect. The DNA Commission of the International Society for Forensic Genetics states that the evaluation of parentage "shall be founded on mutually exclusive hypotheses", and its own example of 3 such hypotheses is that the father is the putative father, the putative father's brother, or a man unrelated to him (Gjertson and colleagues, "ISFG: Recommendations on biostatistics in paternity testing", Forensic Science International: Genetics 1:223–231, 2007).
~ To its credit, the study does weigh Randolph Jefferson as an alternative--but only by simulation:
The Randolph line is represented only by 2 samples at the 6th generation of descent from Randolph Jefferson, not by Randolph Jefferson himself. There is no Randolph genome in this study at generation 0, which is what a direct comparison would require, and both of those descendants carry additional intrafamily marriage in their ancestry. Dr Kurin's team spent years searching for the remains of Randolph Jefferson himself, and then of Peter Field Jefferson, a son of his first marriage. But that first wife, Anne Lewis, was Randolph Jefferson's own first cousin, so that son's DNA--even had it been recoverable--would carry ancestry shared with Thomas Jefferson through both parents, and would be difficult to separate from Thomas Jefferson's own. Dr Kurin was aware of that cousin marriage. He cited it himself to The Monticello Association in July 2026, adding that in DNA research "you want to try to get as clean as you can"--yet his team spent years pursuing the son of that very marriage. My petition in Nashville instead seeks a son of Randolph Jefferson by his second marriage, which involved no such intermarriage, so any Jefferson family DNA he carries came to him through Randolph Jefferson alone.
~ The other candidates were never tested--the Carr nephews:
The Carr nephews, named in Jefferson family accounts for 2 centuries, were not tested either. Those accounts are specific. In a letter to her husband of 24 October 1858, Thomas Jefferson's granddaughter Ellen Randolph Coolidge passed on what she had from her brother, Thomas Jefferson Randolph--that there was "a general impression" that Sally Hemings' 4 children "were all the children of Col. Carr", and that he had heard Peter Carr say, with a laugh, that "the old gentleman had to bear the blame of his and Sam's (Col. Carr) misdeeds" (Ellen Wayles Randolph Coolidge to Joseph Coolidge, 24 October 1858, Ellen Wayles Randolph Coolidge Letterbook, University of Virginia Library). Thomas Jefferson Randolph gave the biographer Henry Randall a related account--when he put a newspaper article on the charge into Peter Carr's hands, Peter wept and said to his brother, "arnt you and I a couple of … pretty fellows to bring this disgrace on poor old uncle who has always fed us!" (Henry S. Randall to James Parton, 1 June 1868, James Parton Papers, Houghton Library, Harvard University). These are family accounts passed on secondhand, and historians have long debated how much weight they deserve. That is exactly why they needed to be tested with DNA rather than set aside. Dr Kurin told me he understood Foster's 1998 study to have eliminated them. It did not--it excluded the Carrs only for Eston Hemings, and the preprint itself says so, acknowledging that Foster's study "did not address the paternity of Sally Hemings's other sons".
~ Madison Hemings had no living male line descendants in 1998 and has none today, so his line could not be tested then and cannot be now:
For Madison Hemings, the preprint instead dismisses a Carr father as "implausible" on the strength of the same segment calls now in question--without any Carr DNA, and without modeling a Carr scenario at all, even by simulation--although it does weigh Randolph Jefferson that way. On the line carrying this study's strongest claim, the Carr question has never been answered by DNA at all. There is a test that would answer it, and it does not require any Carr remains. In my own database, Madison Hemings descendants and descendants of the Carr nephews share 10 to 20 IBD segments. If Thomas Jefferson fathered Madison Hemings, nearly all of those shared segments should also appear in his recovered genome, because his only connection to the Carr nephews runs through the parents he shared with their mother, his sister. If a Carr nephew fathered Madison Hemings, only a minority of them--roughly a quarter--should appear there, the rest coming from the Carrs' non-Jefferson side. On my data the shared segments look more like the second pattern than the first, which is why I continue to regard a Carr nephew as a live possibility rather than an excluded one. I hold that tentatively, and I would revise it on the study's data if it shows otherwise. A second test is broader still--take a segment that 2 descendants must have inherited through Thomas Jefferson, and check that his recovered genome actually carries it. A segment that had to pass through him but is missing from his genome would point to a defect in the genome or in the pipeline that called it. Both tests can be run in a week by anyone holding the data, and neither can be run by anyone outside the team until the data is released.
~ The comparison group is small:
After 15 years, Dr Kurin's conclusion depends on only 12 usable descendant samples across all lines, with as few as 2 in some. Sixteen people were sampled, but 2 yielded no usable DNA, 1 participant withdrew, and 1 was sequenced only for his mitochondrial DNA, leaving 12 in the comparison--8 Hemings descendants at the 5th generation through Madison and Eston, 2 Jefferson descendants at the 5th generation through his daughter Maria, and 2 descendants of his brother Randolph. The participant who withdrew is Dr Mary Brady-Atkins, a 5th generation descendant of Sally Hemings, a PhD, and a retired psychology professor, who gave her reasons to the press at the time. She objected to the order in which the work was being released--"A finding is submitted to a journal. It is reviewed by those an editor selects rather than by those the investigator selects. It is published together with the data supporting it, so that others may test it. Only afterward is it written about for a general audience"--and said of this study that it "has proceeded in the reverse order". She also said she had "not been permitted to read what is being published about my own family, when my sample is among the reasons there is a book to publish" (Stephanie M. Lee, "The Smithsonian Used Her DNA. And Lost Her Trust.", The Chronicle of Higher Education, 17 September 2026).
~ That panel of 12 compares to my 125 samples collected (85 defensibly usable), which include Jefferson, Hemings, and Carr samples:
The reduction to 85 is a matter of study design rather than of attrition. A parent and child sampled in the same household are not two independent observations, so they are not counted twice, and siblings are averaged to a single value so that no one family carries disproportionate weight. Handling closely related samples in this way is ordinary practice in genetic studies. At this generational distance the amount of shared DNA varies enormously from one relative to the next, and a panel of the small size used in Dr Kurin's study cannot appropriately capture the range.
~ The reference is a composite:
The genome was not built from a single lock. The team began with Monticello packet 1956-31, and over the years, in the book's words, the genome was "continually upgraded and elaborated with advanced techniques and the addition of DNA from other locks of Thomas Jefferson hair". The preprint identifies those additions as 2 further packets, 1968-67-7 and 1968-67-8, and traces all 3 to Nicholas Trist, who clipped the hair after Thomas Jefferson's death. The reference is therefore a composite, assembled over several years, and the calls made against it were revised as the software changed. Neither the book nor the preprint says when the genome was assembled, or which version of it produced the published figures.
~ The genome was read only about twice over:
The Jefferson genome was recovered at about 2-fold coverage, meaning each position was read on average just twice, and much of it once or not at all. For comparison, a team including the Max Planck Institute tested 8 independently sourced locks of hair attributed to Ludwig van Beethoven, found that 5 came from a single man and 2 did not match, and sequenced his genome from the authentic locks to 24-fold coverage--work its authors describe as enabling "the sequencing of high-coverage genomes from small quantities of historical hair" (Begg and colleagues, Current Biology 33:1431–1447, 2023). How much DNA survives in old hair varies with the specimen, so the two are not directly comparable. But at 2-fold, the choice of method for finding shared segments carries far more of the weight. The preprint tests the locks genetically--the DNA is male, matches a living relative on Thomas Jefferson's mother's line, matches the Jefferson Y chromosome, and comes from one individual across all the locks. Those checks are real. But every one of those tests is equally true of Randolph Jefferson, who had the same mother and the same Y chromosome. What distinguishes Thomas Jefferson's hair from his brother's is the written provenance of the packets, not the DNA. That is worth pausing on, because provenance belief has already failed once inside this project, as the book itself recounts--the Jefferson Bible hairs were taken to be Thomas Jefferson's for years on the strength of where they were found, and the DNA eventually showed they were not his at all. The same kind of belief is what now distinguishes these 3 packets from any other Jefferson family hair. In fairness, the amount of DNA the acknowledged Maria Jefferson descendants share with the genome is consistent with its being Thomas Jefferson's.
~ My own study takes a different approach:
I have supplied the Max Planck Institute with locks of rootless hair not only from Thomas Jefferson himself, but also from both of his daughters, Martha and Maria, and from 4 of his grandchildren. A daughter shares about half of her DNA with her father but only about a quarter with her uncle, so hair from Thomas Jefferson's own daughters can test genetically, rather than by written provenance alone, whether a genome is his or his brother's. Each daughter inherited half of her DNA from her father, so what Martha and Maria carry from him is Thomas Jefferson's own DNA--not his brother's--and the 4 grandchildren extend that check a generation further. Those samples also allow the genetic data to be checked for consistency across the family. They are not used to fill in the gaps in the recovered genome itself--close relatives are deliberately kept out of that step, because they carry sample-specific effects of their own and would themselves need filling in, which compounds any error. For an individual as recent as Thomas Jefferson, a modern high-quality reference panel does that job well, provided the baseline coverage is sufficient.
~ The Jefferson comparison line was chosen on a mistaken premise:
Dr Kurin sampled acknowledged Jefferson descendants only through Thomas Jefferson's daughter Maria Jefferson Eppes, and not through his daughter Martha Jefferson Randolph. His stated reason was that the Maria side had fewer intrafamily marriages. But in fact the reverse is true. Maria married her own first cousin, John Wayles Eppes, so her descendants inherit John Wayles' ancestry twice over and are related to one another in more than one way. That makes the Eppes line a poor yardstick for how much of Thomas Jefferson's DNA an ordinary descendant carries--which is exactly what a comparison group is for. Martha, who did not marry into the Wayles family, would have been his better choice.
~ These were not large, outbred families:
The statistical model assumes that the DNA a descendant shares with Thomas Jefferson came through the specific family relationship being tested. In these families that is not a safe assumption. The Jefferson, Randolph and Lewis families intermarried more than once--Randolph Jefferson married his own first cousin--and the preprint notes that at least one known intrafamily marriage is left out of its simplified family tree. Shared DNA that arrived by another family route can look, to the model, like DNA inherited from Thomas Jefferson himself. The authors do account for Randolph Jefferson's marriage to his cousin, and they are right that the reported half-sister relationship between Sally Hemings and Thomas Jefferson's wife does not affect these comparisons, since Thomas Jefferson carried no Wayles DNA. What the paper does not show is that every such route is captured in its simulations.
~ The strongest single result came at the very end of their study:
Dr Kurin has described the sequence himself. Writing in Smithsonian magazine, he recounts that after the first Madison Hemings samples the result was inconclusive--"the levels were low enough in those cases that they could have been descendants of Jefferson's brother Randolph--we couldn't tell"--and that the 3 year search for Randolph Jefferson's remains then failed. "We pivoted," he writes, "thinking that perhaps sampling more Madison Hemings descendants might bear fruit," and in early 2024 he collected further samples, which produced the high figures the conclusion depends on ("Did Thomas Jefferson Father Sally Hemings' Children?", Smithsonian magazine, November 2026). The 5.1% figure he has emphasized belongs to one of what his own book calls the "most recently analyzed" Madison Hemings descendants, and he was still accepting new samples only 6 weeks before the announcement. That sequence matters. The chance of drawing at least one unusually high value rises with every additional sample, so a search that continues until a strong result appears, and then stops, tends to produce a stronger result than the evidence as a whole supports. Recognized genealogical standards treat this as a bias to be guarded against--no single DNA result constitutes proof, and a result far above the rest of the pool must be investigated for another explanation rather than accepted as confirmation (Genetic Genealogy Standards 19 and 20; Board for Certification of Genealogists Standards 17 and 43). Continuing to collect samples until a sufficiently high value appears, and then stopping, is the pattern those standards are written to prevent. The objection is therefore to the study's conformity with the applicable professional standards, not necessarily to its arithmetic. The study should state how many rounds were run, how many individuals were tested in each, and whether the reported certainty was adjusted for the number of comparisons.
~ The competing explanations for an excess were not excluded:
Beyond ordinary chance in inheritance, there are 3 main ways a Madison Hemings descendant could carry more Jefferson DNA than an acknowledged Jefferson descendant--(i) an additional line of Jefferson family ancestry entering later in the pedigree; (ii) a false positive rate in the segment caller at 2x coverage; or (iii) bias in the reference population figures used to judge chance matches, applied to modern genomes of mixed ancestry. Wayles ancestry is not one of them, since Thomas Jefferson's genome carries no Wayles DNA. None of the three is excluded in the book, and the preprint does not rule them out. The test is straightforward, and it is one Dr Kurin's team can run--do the 5.1% descendant's excess segments fall at the same chromosomal positions as segments in the Randolph line descendants? If they do, they are Peter Jefferson and Jane Randolph ancestry rather than Thomas Jefferson's own, and roughly half of any Jefferson signal cannot tell the two brothers apart in any event.
~ There is a more basic problem with their study:
The acknowledged Jefferson descendants are the study's positive control--they establish what a true descendant of Thomas Jefferson at that generational distance actually looks like. Here several Hemings descendants score well above them, and the preprint itself says the 2 usable Maria Jefferson descendants fall "somewhat less than" the amount expected. Dr Kurin states the comparison clearly--the descendant with the highest figure, he writes, "shared more DNA with Jefferson--5 percent--than any other person tested for the study, including those descended from his wife, Martha", and he described that figure to her as "overwhelming evidence". When the samples being tested outscore the controls meant to calibrate them, the controls have failed at their job--either they do not represent a true descendant, or the test is measuring something other than descent from Thomas Jefferson. With only 2 controls, the study cannot tell which. The explanations that would account for it--an additional shared line, pedigree collapse, a calibration error, or bias in the reconstructed reference--each have to be excluded before high values can be read as confirmation, and neither the book nor the preprint excludes them.
~ What would settle it:
Release of the recovered genome and the underlying data, so that others can reproduce the findings using the established method, is imperative. So is obtaining the direct comparisons this paternity question has always required--Randolph Jefferson's DNA for Eston Hemings, and the Carr line for Madison Hemings. These are ordinary requests, and Dr Kurin has already said the genome will be deposited, but it has not yet been.
~ What I am not saying:
I am not saying that Thomas Jefferson did not father these children, and I am not saying Dr Kurin's study is wrong. But what I am saying is it has not yet been shown to be right by anyone outside the team that produced it. The families on both sides of this history have waited 200 years for an answer, and they deserve one that stands up under independent scrutiny and analysis. If the data is released and the findings survive independent testing, I will say so clearly.
~~~~~
This link will lead you to the critique above as well as additional links for further study:
https://www.tjheritage.org/
