Genealogy by the Numbers: Confirming a Relationship with DNA

There have been few items in the genealogist’s toolbox that can demolish those ever present brick walls; however, one that I have learned to depend upon to help solve some puzzles is DNA analysis. To date, four personal genealogical brick walls have collapsed through the use of both Y-DNA and autosomal DNA testing. Some of these dead ends could not be settled with almost 35 years of traditional research as documentation did not exist.

Last week, autosomal DNA testing solved a mystery that I have been pondering since I discovered a relative at the Boston Public Library in 1988. Named Ruth Day, this woman could have been my father’s half-sister, first cousin through his Aunt Susie, or first cousin through his Aunt Lydia. While I had a theory on her relationship to my father, I could not confirm my suspicions until DNA analysis indicated a solution to this genealogical quandary. This was confirmed by the numbers in light of other family information.

By comparing the amount of autosomal DNA shared between relatives, it is possible to estimate a relationship. Recombination of autosomes occurs randomly; however, as the amount of DNA shared between individuals lessens, the ability to predict a relationship with some accuracy becomes more difficult. For those who have shared DNA with my own family at under 0.75%, some of the relationships are quite distant. Steve Mount has a blog post that explains this very well.

To date, I have been able to confirm only three heretofore unknown relationships – a seventh cousin; a double 9th cousin, once removed; and a 12th cousin, twice removed. These relationships are quite distant and some share more DNA with me than those with known relationships. So predicting at lower levels is a gamble, but predicting approximate relationships with a higher percentage of DNA is possible.

With that said, two individuals from an endogamous group (such as Ashkenazim) will have a higher shared percentage of DNA with more distant relationships. Even those with Colonial ancestry might have higher shared DNA percentages than indicated and that may explain the three distant relationships mentioned above as all three are related to me through my grandmother’s New England colonial ancestry.

Autosomal DNA Shared Between Close Cousins

With each generation, the average amount of shared DNA decreases by half; however, because recombination is random, this number is not exact. For example, I share at a less than average amount of DNA with one of my brothers – 41.00%. His son and I share only 14.90% of our DNA – a difference of 63.7% less than the preceding generation. With the range is outside of the normal range for a nephew, 23andMe calculates us as first cousins.

Add another generation with my daughters and the percentages tend to right themselves. My youngest daughter and her first cousin share 9.13% of their DNA – a difference of 38.7% from the preceding generation – indicating that a more than average amount of DNA was retained with this next generation – although lower than the average 12.5% estimated share. My other daughter shares 7.38% with her cousin, which is a very low share for a first cousin; however, she retained 50.5% of the DNA that her father and cousin shared.

Following a normal curve, statistically 68.2% of all matches should occur within ±1 standard deviation (σ) from the mean score. The following chart indicates the ranges of ±1 standard deviation from the mean for each degree of relationship from first to third cousins.

Since there are a variety of relationships that fall within these ranges, I have identified the percentages according to the degree of relationship between the two subjects. These are based on the civil definition of degrees of relationship which is calculated by adding the total distance from the common ancestors for both subjects. Half first cousins have the same civil degree of relationship (4) as full first cousins; however, the amount of DNA shared is less and appears to be the same as first cousins, once removed (a degree of relationship of 5).

Relationship	DNA Degree   of Relationship
First Cousin	4
First Cousin, Once Removed	5
Half Cousin*
Second Cousin	6
First Cousin, Twice Removed
Half Cousin, Once Removed*
Second Cousin, Once Removed	7
First Cousin, Thrice Removed
Half Second Cousin*
Half Cousin, Twice Removed*
Third Cousin	8
Second Cousin, Twice Removed
Half Second Cousin, Once Removed*
Half Cousin, Thrice Removed*

*half cousins have a degree of relationship one step higher; however, they share DNA with relationships of a lower degree.

Finding Ruth Day

During spring 1988, I took a marathon research trip that took me to Newark and East Orange, New Jersey; Boston, Massachusetts; Concord and Claremont, New Hampshire; and Pittsburgh, Pennsylvania. During that journey, I had the opportunity to visit the Boston Public Library and was able to find my great-grandfather (Thomas Day) in the soundex for the 1880 and 1900 census records for New London County, Connecticut.

In the 1900 census, I discovered that his household contained a nine-year old granddaughter named Ruth Day. This was the first time that I encountered this name and really didn’t know how she was Thomas’ granddaughter. Since Thomas had three daughters and no sons, there were three primary possibilities of how she could have been related to me.

Although I pondered this over the years, I didn’t think much about it until I visited the National Archives in July 2009. During this trip, I decided to have Thomas Day’s Civil War pension record pulled for a second look. I had accessed it in February 2000 and it provided me very critical information that helped me enlarge my family tree. While looking at the documents, I noticed that Ruth Day was utilized by her grandfather as a witness to his documentation. The final entry that she witnessed was on March 9, 1907. Her address was listed as being North Stonington, CT.

Not being able to find her under Ruth Day in the 1910 census, I assumed one of three scenarios: she had relocated, got married, or had died. I assumed that she must have married. I repeated my search to find a married woman by the name of Ruth who was born from 1889-1891 and who was living in North Stonington, CT.

Three possible married women named Ruth from New London County were returned in the search on Ancestry. While one lived in Lyme and another lived in East Lyme, the third was living in Stonington. Ruth Tillinghast, the wife of Walter B. Tillinghast, seemed like a prime suspect to be my Ruth Day. I searched her husband’s name and found him among the family trees posted on Ancestry with his wife “Ruth Day.” In addition, I learned about the story of their son Julius who, at the age of 4, was adopted out because his mother never bonded with him. Julius Tillinghast became J. Russell “Russ” Deckard.

Russ Deckard Chief Radio Officer, SS Evangeline, 1941, World War II

The tree was managed by Deckard’s son Gerald, and he included the transcript of the letter sent by Russ’ birth father to his adoptive father.

Mr. Deckard,
I had no trouble getting Mrs. Tillinghast to sign the paper. I saw her about an hour and she did not mention his name even how he was or where he was or who was to adopt him. It makes me feel bad the interest she has shown in him but it is God’s will. I believe he knows best. You need not worry about her ever trying to see him for she won’t know where he is or who has him. I myself feel bad to do this but think it is the best for him as his mother takes no interest in him. I shall not try to see him here after. I was in the insurance office at Westerly but they had sent my book Saturday morning to the city so you will have to go the insurance office here in the city have it changed for yourself for I may not get the book before I go away but if I do and have time I will make them fix the book and policy. The next premium will be due the 20th of May next as I paid it in advance yesterday 9 weeks. I would like it to have paid 1 year in advance but could not on account of money. I am to have no money for two months in Kavey (???) as they would hold one month back.

P.S. Mr. and Mrs. Deckard. I can never thank you enough for this kind and charitable deed you have done for me and mine and hope you people never have cause to regret it. I also hope you have no trouble to train him to your likes and dislikes.

I remain your respectfully Walter B. Tillinghast

Additional searching provided me further information regarding Ruth Day including her death at 45 in Rochester, Minnesota. I purchased her death certificate online and it arrived a few days later. Her tragic end was detailed by the attending physician. It appears that Ruth fractured her right femur during a fall in 1933. In late January 1935, she had a bone graft operation at the Mayo Clinic because the break was not joining together; however, she developed a pulmonary embolism following her operation and died. In time, her husband remarried.

Even though her approximate birth date and location were known, her birth did not appear to be registered. I could not be certain on which one of the Day sisters was Ruth’s mother. Although I knew more about her in 2009, my ability to determine the identity of her mother had not changed since I discovered her existence 21 years earlier.

Daughter of Susie Eva Day?

As I analyzed the data, her birth on December 22, 1890 at Gales Ferry, CT, and other data helped narrow the possibilities. It would be unlikely that she was Susie’s child, as Susie was married at the time and would have been pregnant with her second child Myrtle in December 1890.

Susie and John Merriman with their youngest child, Lawrence Raymond Merriman

Myrtle Merriman was born in May 1891. If Myrtle was premature, it is possible (but highly unlikely) that Susie could have given birth to both girls in that short period; however, there is also question regarding Ruth’s birth year. The 1900 census lists her being born in December 1890; she is also listed with the following ages on her last birthday during the following census: 1900 – 9, 1910 – 19, and 1920 – 29 – all corroborate an 1890 birth year.

The 1930 census and her death certificate, however, indicate she was born in 1891. If this were the case, her birth 7 months after Myrtle’s could be believable, but not very probable. With Ruth having the name of Day, it is another strike against her being Susie’s daughter as she was already married to John Merriman at the time of Ruth’s birth.

Daughter of Lora Gardner Day?

Although Susie could be eliminated by the virtue of her own pregnancy, this was not the case with the second daughter Lora – my grandmother. In my mind, she was eliminated because in 1904 she gave birth to the fifth of her sixth children and named this daughter as Ruth E. Dalton. While it is possible to have two or more children with the same name (ask George Foreman), it wasn’t very probable.

Lora with her youngest child: Charles Ellsworth Owston

Daughter of Lydia D. Day?

As far as I was concerned, Lydia, the youngest daughter of Thomas and Amy Day, was the best candidate for being Ruth Day’s mother. Lydia was born on September 10, 1875 and she would have been 15 years old when Ruth was born. Although younger than her sisters, circumstantial evidence pointed toward Lydia; however, I had no proof that she was the rightful mother of Ruth Day.

On Ruth’s death certificate, a mother isn’t listed; however, Thomas Day was listed as her father. A birth certificate for Ruth Day was not found at Ledyard which includes the village of Gales Ferry. Although Thomas Day is listed on the death certificate for Ruth, it does not appear that he was her actual father. In the 1900 census, she is listed as his granddaughter and in various documents in his Civil War pension file, she is not listed among his children.

Lydia (left) with sister Lora; identity of the baby is not known

Of the sisters, Lydia is the one of which I know the least. She was the first of the three to pass away (August 1931) and she is the only one of the three to divorce a husband. She married Frank Ernest Botham circa 1892. One child, Elmer Ernest Botham, was produced from the union in 1893. It is uncertain when she divorced, as in an affidavit dated April 14, 1898, her father lists her as Lydia Botham and that she was married. In the 1900 census, Frank Botham was listed as still being married and having been so for eight years – even though Lydia was married to someone else by then.

When she married William Henry Gorton in 1898, it appears that custody of her eldest son was retained by his father. The relationship she had, if any, with Elmer Botham is currently unknown; however, it appears to be negligible at best. He is not counted among her children in the 1900 and 1910 censuses. Four children were produced from her second marriage and included William H. Gorton, Jr. (1899), Henry Reynolds Gorton (1902), Mary Gorton (1904), and Elmore Gorton (1907). It appears that young Elmore died prior to the enumeration of the 1910 census.

Being the youngest child of Thomas and Amy Day, Lydia was not quite 13 years of age when her mother passed away. During her critical adolescent years, a mother’s influence was absent, and this may have had a profound effect on Lydia. If she were Ruth’s mother, her pregnancy occurred in less than two years from the time of her mother’s death.

Since the rearing of Ruth was consigned to her grandfather and Lydia’s child Elmer was raised by his father, it creates an interesting picture of Lydia’s psychosocial development. This same abandonment scenario appears to have occurred with Ruth and her son Julius and therefore, the behavior could be considered cyclic. That is, if Lydia was her mother – but neither Susie nor Lora had similar experiences with their children. In fact, Susie took in a grandchild and raised him as her own.

Unfortunately, psychosocial similarities are not accepted proofs of descent. Without a birth certificate to prove Ruth Day’s maternity, I was back to square one – I knew she was related but not how. In August 2009, I messaged Gerald Deckard through his Find A Grave account to attempt to make contact with this newly found cousin; however, he did not discover my message until January 2012. After a series of emails, we decided to try the autosomal DNA route to see if a specific relationship might be confirmed.

Shared DNA Segments

By using induction, we should be able to confirm the relationship between Gerald and others who are descended from Thomas Wesley Day and Amy Alice Champlin. The amount of shared DNA among the subjects should be indicative of their relationships. In addition, the percentage of shared DNA should fall within ±1 standard deviation of the statistical mean for that relationship.

Although I do not have contact with any known descendants of Lydia, we do have DNA results from seven descendants of Lora and two descendants of Susie. For the purpose of this analysis, we will consider the five closest matches. If Ruth were the child of Susie or Lora, then the relationship between Gerald and the appropriate subject would be that of a half cousin, once removed.

The amount of DNA shared by half cousins, once removed would average at 3.13%. The range within ±1 standard deviation would be 2.06% to 4.20 %. Extending that range to ±2 standard deviations would increase the range from 1.63% to 4.62%. Theoretically, 95.4% of all matches within this close relationship should fall within this range. Even at ±3 standard deviations, the percentage shared ranges expand to 1.57% to 4.69%. Nearly all matches (99.7%) would fall within ±3 standard deviations.

If Ruth were descended from Lydia, Gerald’s relationship with the others would be diminished to that of a second cousin, once removed. The amount shared for this relationship averages at 1.56%. The ranges for this relationship would be 1.03% to 2.10% ±1 standard deviation and a range of 0.82% to 2.30% with ±2 standard deviations.

Subject	Grandmother	Amount of   Shared DNA
Jim	Lora	1.47%
Chuck	Lora	1.29%
Suzanne	Susie	1.17%
George	Lora	1.15%
John	Lora	0.92%

Although more samples from Susie’s line at the grandchild level and samples from Lydia’s known descendants (if any exist) would have been optimum in confirming Ruth Day’s maternal ancestry; however, by combining the information we currently know with the amounts of shared DNA, we may be able to infer that Ruth’s mother was Lydia.

Four of the five subjects are grandchildren from Lora and three of these subjects' results fall within ±1 standard deviation. A fourth, John, falls within ±2 standard deviations. There are enough samples here to confirm that Ruth Day was not the daughter of Lora Gardner Day and that her own daughter Ruth E. Dalton was probably named for this cousin.

As for Susie’s descendants, we only have one sample from a grandchild; however, since that number falls within ±1 standard deviation for a second cousin, once removed; it is likely that the suspected relationship is accurate. Suzanne falls completely outside of the realm of a half-cousin, once removed even with ±3 standard deviations from the mean for the higher relationship. Even if she were an outlier, it would be expected that that the numbers would be higher for a supposed closer relationship. Two of Lora's grandchildren have higher results.

As mentioned earlier, there are four other individuals that have been tested at 23andMe from this family: three great-grandchildren of Lora and a second great-grandson of Susie. The descendents of Lora should be related to Gerald as third cousins and Susie’s great-great-grandson, Craig, should have a relationship as a third cousin, once removed.

Subject	Ancestor	Amount of  Shared DNA
Lora	Lora	0.81%
Craig	Susie	0.79%
Kristen	Lora	0.63%
Michael	Lora	0.00%

As more distance is placed within a specific relationship, it is my experience is that it becomes more difficult to predict the suspected relationship. The average estimated DNA shared at the third cousin level is 0.78% while a third cousin, once removed would be 0.39%. It is also estimated that only 90% of third cousins will have shared DNA. That percentage drops to 0.45% at the fourth cousin level.

In the case of these family members, Michael shared no DNA with his suspected third cousin – but among the previous generation, Michael’s father had the least amount of DNA shared with Gerald. Lora and Kristen both fell within a ±1 standard deviation; however, Craig shared nearly twice as much for a third cousin, once removed. His results appear to be like that of a third cousin. He also has five shared segments.

Conclusion

Although Craig’s unusually large match might implicate Susie as Ruth Day’s mother, other factors include the amount of DNA shared between Suzanne and Gerald and Susie’s pregnancy with Myrtle that overlaps the birth of Ruth.  This evidence appears to eliminate the slightest possibility that Susie was Ruth’s mother.

With four grandchildren of Lora sharing with Gerald at a relationship consistent with second cousins, once removed, she is eliminated by virtue of the lower numbers. The fact she also had named her daughter Ruth in 1904 adds weight to this argument.

Although samples from Lydia’s line and additional samples from Susie’s grandchildren would provide further data and make the case stronger, I believe by combining the circumstantial evidence with the amount of DNA shared suggests that Lydia was Ruth Day’s mother. I have always assumed this, but had an inability to confirm my hypothesis.

DNA analysis in this regard helped break down a brick wall in our family tree; however, enough other data was present to lead to this confirmation. Some close DNA matches, however, may not be able to suggest a relationship without the presence of other data. A suggested relationship by the numbers, however, may provide an impetus to look for the relationship in the correct direction; and thus, shared DNA can add to the body of genealogical evidence in determining one’s relationship with another individual.