Genetic Tests

It was correctly alleged by Jenny Jerome's sisters that the father of John Spencer-Churchill was not Randolph Spencer-Churchill, but Evelyn Boscawen (Anne Sebba, American Jennie: The Remarkable Life of Lady Randolph Churchill, Norton, 2008; Anne mentioned in her conclusion the desire to pursue a genetic analysis). The evidence for these claims has been examined in detail by Celia and John Lee's books (The Churchills - A Family Portrait (New York: Palgrave Macmillan, 2010), Winston & Jack - The Churchill Brothers (London: Mackay, 2007)). The Lee's concludes that much of the evidence for John's non-Spencer Churchill paternity was a combination of political libel by specific individuals, and journalistic exaggeration of Jenny Jerome's life. A more accurate assessment is by Mary S. Lovell (The Churchills: A Family at the Heart of History - from the Duke of Marlborough to Winston Churchill (New York: W.W. Norton, 2011)), who makes the observation that John Spencer-Churchill had a closer resemblance to some of Winston's children than Winston himself, and in turn the same can be said that Johnny Churchill resembled Winston more than John.

In 2023, to account for paternal relatives with a genetic proximity of 50 Cm, but which did not fit into the family tree, I ran a series of simulations using ancestry.ca's "link" feature. There consequently emerged overwhelming evidence that Johny Spencer-Churchill's father was not Randolph Churchill, but instead Evelyn Boscawen, exactly as alleged by Jennie Jerome's sisters. I tested the hypothesis by changing the composition of the family tree and seeing how many unexplained relatives fell into a spot on the tree, and over a dozen relations in the expected Cm range of 6 to 20 from the Boscawens were now linked. The problem with the measurement process is that there remain strong Cm links with persons descended from several generations of Spencer-Churchills, which confounded earlier tests, simply because the British aristocracy is closely related.

The discussion below was the original test of a number of paternity hypotheses, preserved here to demonstrate the process, but is since been superseded.  

The essential sequence of events, as reconstructed from various accounts listed above and in the accompanying bibliography at the end of the website, is that Winston Churchill was born at Blenheim in 1875. Following a scandal involving Randolph Churchill's brother, the family relocated to Dublin, Ireland. Jennie Jerome would hunt frequently, which offered her plenty of opportunities for carousing, although letters between her and Randolph Churchill indicate continued genuine intimacy. Randolph frequented between Dublin and his Parliamentary duties in London. John Spencer-Churchill was born in 1880. In 1885, Randolph Churchill is thought to have contracted syphilis from a Parisian courtesan, There are alternate accounts, less reliable, that Randolph contracted syphilis while at Oxford University, or with a maid at Blenheim, both before the birth of Winston Churchill. Jennie Jerome did not contract syphilis, falsifying those earlier accounts, which is  good indicator that they actually ceased intimate relations shortly after Randolph's self-diagnosis of syphilis.

The genetic evidence is that John Spencer-Churchill is certainly related to Randolph Churchill, son of the 7th Duke of Marlborough. M Schofield / Spencer-Churchill (grandson of John Spencer-Chuchill) has an 8 Cm link to a descendent ("A. M-G.") from a common ancestor, George Spencer, the 4th Duke of Marlborough. "A. M-G." (1954-) is a sixth cousin once removed from Julian Spencer-Churchill (son of M Schofield / Spencer-Churchill). Put simply: to show that John Spencer-Churchill was the son of Randolph Churchill, we need to find a genetic link between a descendant of John Spencer-Churchill and a descendent from an ancestor of Randolph Churchill. A genetic link between the descendants of Randolph Churchill is not optimal, since this link could be traced as common descent from Jennie Jerome, although the strength of this link may still be tested. We then need to ensure that the genetic link is not the result of an alternative non-Spencer-Churchill family link, and this is done by mapping out both family trees within autosomal detection range, which is approximately 400 years. Finally, the proposed alternative fathers of John Spencer-Churchill need to be tested, by seeking genetic links with their descendants, and where these expected genetic links are absent, making a probability estimate of a false positive result of a non-finding. All of these steps are demonstrated below.

"A. M-G." 's family pedigree was mapped out (from ancestry.ca, geni.com, and thepeerage.com) to control for alternate genetic paths, and none were found, although there are links to a dozen shared noble families reaching to the Middle Ages (Pennant, Oliphant, Lascelles, Swinton, Butler, Bedell, Huntly, Villiers, Egerton, Bagot, Molyneaux, Darcy), which are typically beyond the detection range of autosomal techniques. The expected Cm value is 0.75 Cm (assuming step level values with a 40% Cm value between siblings), which is within a magnitude range of 8 Cm. 38% of detected individuals with a genetic link of 8 Cm are likely to be false positives, in the absence of a family tree link. However, the probability of an 8 Cm genetic link with someone expected to be within the range of 8th cousins, so that the 8 Cm genetic link is the result not of shared DNA, but a random matching coincidence, is statistically remote.

What follow are auxilliary attempts to examine the question of John Spencer-Churchill's paternity in light of genetic evidence from two DNA sites. The main problem with both tests is that they are conducted with living descendants of candidates, all of whom may have been from cuckolded lineages. There are a number of persons that we are related to genetically, but whose extensive family trees demonstrate no common ancestor, clearly due to error or ancestral deception (for example Anne Hamilton). Cuckoldry among the nobility in the 19th Century is assumed in the academic literature to be lower than the rate among the general population, about 2.5% generally in the British population. Nevertheless, parents of the British nobility married their children strategically for money and power, and children were therefore instrumental and often neglected, poorly parented and therefore poorly behaved, and this cycle was efficiently reproduced, but giving little incentive to mothers to remain faithful to their marriage vows.

There are a number of methodological issues. The first is that members of the English peerage generally do not submit their DNA for tests on recreational ancestry sites, unless they have reason to suspect their ancestry, or are accessing health, ancient haplogroup, or ethnicity features, as are available on 23&Me. Most members of the nobility have easy access to extensive records of their ancestors at their home, through Burke's peerage, or through thepeerage.com. Furthermore, propertied peers do not typically wish to risk a loss of their wealth, now that genetic principles of relations can be used to determine property rights in Britain. The consequence is that nobles tend to be significantly under-represented on these websites, and the family pedigrees must be mapped in order to correctly identify members of the family that no longer carry the paternal family name. The second issue is that large segments of the British nobility are interrelated (cosanguine) within 650 years, which is the oldest surviving bloc of common chromosomal commonality I have found. Given that the vast majority of relatives, associated at 20 Cm or less, are likely related through blocs of DNA that demonstrate these disproportionately strong associations, it becomes almost impossible to disentangle with certainty, the precise common ancestor.

Genetic Test #2 of John Spencer-Churchill's Paternity

A genetic comparison made on ancestry.com and 23&Me between ancestors of the principal candidates for the paternity of John Spencer-Churchill are presented here. The finding of a genetic relationship between M (John Spencer-Churchill's grandson) and J.M., through their common ancestor of the 7th Duke of Marlborough, indicates that Jenny Jerome and a relative of Randolph Churchill are the parents of John Spencer-Churchill. The evidence also rejects alternative parentages for John Spencer-Churchill. There are four alternative principal candidates, in the popular literature (cited above), for John Churchill's paternity: Edgar Vincent, Evelyn Boscawen, John Strange Evelyn, and John Thornton Rogers. Each of their pedigrees, showing known descendants, is in an excel file at the bottom of this page.

Three descendants of Evelyn Boscawen[4][5] (all 2nd cousins once removed at their closest) were found to have no genetic relationship with the descendants of John. There is a 0.1% likelihood of a false negative at the position of 2nd cousin once removed (one individual), and a 2.3% likelihood of a false negative for a 3rd cousin (which applies to the descendants of M) [6]. This would be a conclusive rejection of Evelyn Boscawen as a candidate, and it demonstrates the extreme variance in conclusions when dealing with small sample sizes.

Two descendants of the cousins of John Strange Jocelyn[7] (his line ended within three generations), whose closest position (through his father Robert Jocelyn 1788-1870) was 3rd cousin twice removed, were also not found related. There is a 30.7% likelihood of a false negative for a relationship with M, between 30.7% and 69.8% likelihood of a false negative for a relationship with M's children, and between 80-90% likelihood of a false negative for a relationship with M's grandchildren (albeit these results are not independent). There is substantial narrative evidence that Jocelyn is not a likely paternity candidate because of his general absence, his age difference with Jennie, and because he was a good family friend of Randolph Churchill's father; John Spencer-Churchill's middle name was "Strange", which has been taken by some observers are as an indicator of paternity, although it is more obviously explained by the fact that he was John Spencer-Churchill's god-parent.

A fourth candidate, suggested by genetic evidence, is John Thornton Rogers (1834-1900).[8] The nature of the evidence is a large independent cluster of interrelated relatives of M. Spencer-Churchill (father of Julian Spencer-Churchill) with a strength of 50 Cm. He was a frequent visitor to Marlfield House outside Dublin (now a resort), and in 1879-1880, at the time of John's conception, estranged from his wife. He is a plausible individual that Jennie Jerome would have encountered on her hunts during her stay in Ireland, although his recreational habits are unknown. Two of John Roger's descendants (C.B.D.) related genetically to three of John Churchill's descendants, with Centimorgan values in the expected range of second cousins. However, the strength of the genetic evidence is undermined by the complete absence of a relationship with three of the other of John Churchill's descendants, which has only a 2.3% likelihood of being a false negative. Although John Rogers' descendants clustered in near isolation from other lineages (with a cut-off of 20 Cm), there remained confounding effects from multiple common ancestors at least two centuries prior. John Thornton Rogers was selected for testing because of his link to relations with strong CM values, but in fact these genetic links are accounted for by Rogers' descendants' links to the Boscawen family.

Based on these test results, it is extremely unlikely that Boscawen, and given cumulative probabilities (albeit not independent), unlikely that Jocelyn Strange, were John's father.

A final test was conducted to control for the confounding common ancestry of Jennie Jerome, by comparing the autosomal DNA of the six descendants of John Churchill (albeit not independent) with five descendants of the 7th Duke of Marlborough 1822-1883 (who are not also descendants of Randolph Churchill), John Spencer-Churchill's grandfather. These include members of the A., M., and d.l.V., families. These relationships ranged between 3rd cousins to 3rd cousins once removed. One Cm relationship was found, with J.M. . According to data from Blaine DNA, Shared Cm Project 2017, DNA Detectives, FamilySearch.org, zero scores are probable at this range. However, DNA Detectives estimates that 10% of 3rd cousins will be undetected. The P(-Cm) reports more detailed estimates of a false negative, which are low. The results here indicate not that there is a 10% likelihood of non-detection of 3rd cousins, but rather an 80% likelihood of non-detection. C.F.M. and J.M., who are siblings, have the same likelihood of demonstrating a link to M. Spencer-Churchill, but do not. A tree was built for J.M. to search for confounding links, particularly through Dormer or Bertie lineages, but these were not found, except at time distances of 500 years in the shared Scottish genealogy. The low likelihood of a false positive given that the corroboration of genetic and family tree with regard to J.M. and M Spencer-Churchill, indicates a clear link, and rejects the hypothesis of a non-association. The totality of the evidence is that they fail to reject the genetic link between John Spencer-Churchill and the 7th Duke. The 8 Cm genetic link between "A. M-G." (1954-) demonstrates that there is a high probability, at ranges of 3rd cousins and beyond, of there being a detected genetic link with cousins at the range of the 4th Duke of Marlborough.

Although John Churchill's (1650-1722) X-Chromosome haplogroup is available (source unverified though), R1b-M260, given the absence of a surviving male heir, this haplogroup did not match the one available for one of John Churchill's descendants. This is moderately suggestive that John Churchill is not a descendant of Randolph Churchill, although the likelihood that a true cousin is found to be unrelated is 10.3% at 3rd cousins and 54.1% at 4th cousins [9]. The analysis was unsuccessful in subjecting the autosomal links between the descendants of Winston Churchill with those of the descendants of the 7th Duke of Marlborough, as a further control, because none of these descendants communicated to me their autosomal strength with these distant cousins. If the descendants of Winston Churchill are themselves not related to the descendants of the 7th Duke, then there is another problem entirely of establishing a genetic link past five or six generations, the usual limit of autosomal DNA linking.

• [1] All of the individuals referred to were self-identified on public fora
• [2] · https://www.thepeerage.com/p23310.htm#i233094 Retrieved Feb 10 2020
• [3] · https://www.ancestry.ca; https://you.23andme.com/; https://www.familytreedna.com; https://www.myheritage.com; https://www.gedmatch.com
• [4] · https://en.wikipedia.org/wiki/Evelyn_Boscawen,_7th_Viscount_Falmouth
• [5] · https://www.thepeerage.com/p5033.htm#i50323 Retrieved Feb 10 2020
• [6] · https://isogg.org/wiki/Cousin_statistics
• [7] · https://www.thepeerage.com/p3377.htm#i33767 Retrieved on Feb 10 2020
• [8] · https://www.geni.com/people/John-Rogers/6000000009770078866 Retrieved on Feb 10 2020; The peerage.com reference is in error: https://www.thepeerage.com/p26551.htm#i265508 Retrieved on Feb 10 2020
• [9] https://faculty.washington.edu/eathomp/Anonftp/Papers/tpb83_kp_donnelly.pdf

Genetic Test #3 Estimates of Iroquois Ancestry

I examined the indigenous DNA in Julian and his father M, to determine the plausibility of the Churchill family story that the line descends from Iroquois ancestry through Anna Wilcox (1761-1813), an ancestor of Jennie Jerome. A battery of tests for indigenous ancestry available on GedMatch were applied (available in a file entitled "Summary of Indigenous Test Results" and "Indigenous Amerindian" posted below), and most of the observed values fit with the predicted values. A few of the tests found no indigenous results at all in M, but in these instances, the tests were using the Amerindian category as a control rather than as a category of interest.

Julian's DNA is approximately 2/3 similar to M's DNA with respect to segments that derive from M's paternal line. Consequently, M's North American indigenous genetic percentage indicators are 50% higher, rather than double, of Julian's values. The values being in the correct direction, and M's values confirming that the genetic measures for Julian's Hungarian ancestry (the Siberian and East Asian components are sometimes confounded with Amerindian genetics) are good controls, demonstrates that M does indeed have North American indigenous DNA.

However, credible genealogical evidence from 1951 indicates that Anna Wilcox had non-indigenous ancestry, so the story of indigenous ancestry may have been a warping of an earlier account, and there may be another source for the indigenous ancestry. This is problematic because every post-Colombian ancestor of Jennie Jerome is accounted for as having European ancestry, and every post-Colombian American ancestor of the Berties are also accounted for (some European Bertie ancestral lines do not extend to the pre-Colombian period), meaning that the indigenous ancestry, which must have a source, has been concealed. If Anna Wilcox was illegitimate because of her mother, Experience Martin, then Anna would likely be of Mikmaq ancestry, as she was born in Nova Scotia. Her father, Joseph Wilcox, served in a ranger unit composed mostly of Mohawk warriors deployed at Fort Sackville, Nova Scotia. So if Experience Martin had an illegitimate affair at this later period, Anna could be Iroquois-Mohawk. If Anna was herself involved in an illegitimate affair while living in Palmyra, New York state, then her offspring, Clarissa Hall, would be Cayuga-Iroquois. 

Genetic Sources

The following websites are useful genetic resources with regard to this family. The original family tree of 31,221 individuals was developed at, and exported from, ancestry.com, and differs from the family tree excel file on the first page of this website in two respects. First, it is primarily focused on the last 700 years, so it is more thorough in terms of recent family, and secondly, it contains a far more liberal interpretation of connections through illegitimate children, particularly to the royal Stuart dynasty. I have erased these links where I have found them, given the prevailing evidence against them. This family does not descend from the English Stuarts. None of these sites contain Spencer-Churchill, Bertie, or Dormer participants, who are confined to participating on 23&Me and ancestry.com, but they do have large cohorts of Jeromes.

www.geni.com: contains a family tree of 10,000 individuals, that is linked to all of the other family trees on the website, forming a network of over 140 million people. Geni.com has a fascinating feature where it will search and display a path of ancestors between any two individuals in this network.

www.gedmatch.com: contains the raw genetic file: WH6130853 that may be accessed for comparison with other files, with a battery of sophisticated ethnicity estimator software.

www.myheritage.com: contains access to the raw genetic file, and an efficient cluster feature which automatically groups persons together by autosomal genetic similarity. 

www.familytreedna.com: contains a family tree of 10,000 of the most recent ancestors, as well as the most efficient visual chromosomal display for identifying genetic comparisons.