Of Fathers and Sons and DNA
August 30, 2012 11:17 AM Subscribe
My father's grandfather was adopted with very little paperwork or evidence trail as to his biological origins. Earlier this year, my father's uncle got his DNA tested. Unfortunately, the closest Y-DNA relative we've isolated from the tests is 13 generations back. But, I have a related question: All of the matches, through multiple services (having manually entered the data into other services) have a wide variety of non-repeating last names. Given that Y-DNA matches father-to-son, wouldn't there be at least some names repeating?
I've been trying to work through this in my mind, and I can isolate only a few instances in western society where the last name from father to son may differ: 1) adoption, 2) cultural (son takes a traditional reworking of the father's or family's name), 3) Immigration (named after a trade or place of origin), 4) random circumstance.
Given that a generation is, approximately, 20-25 years, I would imagine that our common male ancestor lived approximately 160 to 225 years ago.
Of those matches that are 13 generations back, I've collected 4 different surnames. I've tried to pull genealogical records on those people to see if there is something about their location or heritage that may be familiar to my research, or to each other, and I've hit a big fat zero, zilch, nada.
I've just sent in my DNA test last week, and I'm eagerly waiting the results to see if there is any clarity offered by having two males of the same line tested. But, given the complete lack of meaningful results so far, I don't have very high hopes.
My question focuses on this: At what rate, statistically speaking, should I expect the reasonable spread of last names along the Y-DNA line going back X number of generations?
I've been trying to work through this in my mind, and I can isolate only a few instances in western society where the last name from father to son may differ: 1) adoption, 2) cultural (son takes a traditional reworking of the father's or family's name), 3) Immigration (named after a trade or place of origin), 4) random circumstance.
Given that a generation is, approximately, 20-25 years, I would imagine that our common male ancestor lived approximately 160 to 225 years ago.
Of those matches that are 13 generations back, I've collected 4 different surnames. I've tried to pull genealogical records on those people to see if there is something about their location or heritage that may be familiar to my research, or to each other, and I've hit a big fat zero, zilch, nada.
I've just sent in my DNA test last week, and I'm eagerly waiting the results to see if there is any clarity offered by having two males of the same line tested. But, given the complete lack of meaningful results so far, I don't have very high hopes.
My question focuses on this: At what rate, statistically speaking, should I expect the reasonable spread of last names along the Y-DNA line going back X number of generations?
Given that a generation is, approximately, 20-25 years, I would imagine that our common male ancestor lived approximately 160 to 225 years ago.
Eh? For 13 generations, this would be 260 to 325 years ago. And actually, there's a good chance many of your ancestors in that line were born to men who were 30 and over. (Don't forget that while men may have had their *first* child at a younger average age than today, they kept having children throughout their adulthoods). So that already pushes it out to 390 years ago, possibly.
So, we're talking between 1752 and 1622 just based on those guesses. And in those years, many people in many different places just did not have last names at all. We don't know which culture or ethnic group your ancestor was a part of, and it may have been many years still before that group adopted last names.
posted by cairdeas at 11:33 AM on August 30, 2012 [5 favorites]
Eh? For 13 generations, this would be 260 to 325 years ago. And actually, there's a good chance many of your ancestors in that line were born to men who were 30 and over. (Don't forget that while men may have had their *first* child at a younger average age than today, they kept having children throughout their adulthoods). So that already pushes it out to 390 years ago, possibly.
So, we're talking between 1752 and 1622 just based on those guesses. And in those years, many people in many different places just did not have last names at all. We don't know which culture or ethnic group your ancestor was a part of, and it may have been many years still before that group adopted last names.
posted by cairdeas at 11:33 AM on August 30, 2012 [5 favorites]
Response by poster: Yes, sorry for the typo, there.
posted by thanotopsis at 11:39 AM on August 30, 2012
posted by thanotopsis at 11:39 AM on August 30, 2012
A father may give his son a name, but not a Y chromosome. If you *wink* get what I mean.
We like to think that people in the past were more chaste and faithful persons when in fact they were just as sex-crazed and bed hopping as people today.
A wife taking a lover, especially in an age before birth control, seems to be the most logical reason for this discrepancy. The son of such a coupling would not have the "correct" last name attached to his Y chromosome.
posted by fontophilic at 11:52 AM on August 30, 2012 [5 favorites]
We like to think that people in the past were more chaste and faithful persons when in fact they were just as sex-crazed and bed hopping as people today.
A wife taking a lover, especially in an age before birth control, seems to be the most logical reason for this discrepancy. The son of such a coupling would not have the "correct" last name attached to his Y chromosome.
posted by fontophilic at 11:52 AM on August 30, 2012 [5 favorites]
How do you know that the relations are 13 generations back? If you know all of your forbears from now til then, then surely you know how they're related? Or is this just an estimate from the testing company?
Also, for 13 generations you may be looking as early as 1500, and depending on the culture/country that's too early for fixed lastnames.
posted by Jehan at 11:58 AM on August 30, 2012
Also, for 13 generations you may be looking as early as 1500, and depending on the culture/country that's too early for fixed lastnames.
posted by Jehan at 11:58 AM on August 30, 2012
Also, how much of a match is a "match"? Y DNA tests have different amounts of fineness depending on how many different areas they test. A less fine test will come up with far more matches than a more comprehensive one. A "match" may well be somebody not significantly related to you.
posted by Jehan at 12:04 PM on August 30, 2012
posted by Jehan at 12:04 PM on August 30, 2012
Response by poster: How do you know that the relations are 13 generations back?
Genetics math. 50% of my DNA gets shoved into my son when he's born.
If you know all of your forbears from now til then, then surely you know how they're related
But, I don't. If you read the first line of my OP... Hence the point of getting my DNA tested in the first place.
Also, how much of a match is a "match"?
According to the company, it shows me the differences, and there's only one "difference" on my marker from theirs.
posted by thanotopsis at 12:11 PM on August 30, 2012
Genetics math. 50% of my DNA gets shoved into my son when he's born.
If you know all of your forbears from now til then, then surely you know how they're related
But, I don't. If you read the first line of my OP... Hence the point of getting my DNA tested in the first place.
Also, how much of a match is a "match"?
According to the company, it shows me the differences, and there's only one "difference" on my marker from theirs.
posted by thanotopsis at 12:11 PM on August 30, 2012
There may also just plain not be too many people who've submitted their own DNA to the company you're using, for them to match to. The whole idea of dna-based geneological research is still kind of woo-woo to many people, where they're not really sure what it would tell you, so they haven't added themselves to the database yet. And if they're not there for you to be matched to, you won't match them.
posted by EmpressCallipygos at 12:14 PM on August 30, 2012 [1 favorite]
posted by EmpressCallipygos at 12:14 PM on August 30, 2012 [1 favorite]
Are the last names close? Are the names from the same part of the world? Same ethnic group? Or a totally random sampling?
In 1622, a lot of my ancestors would not have spelled their surnames in a consistent way, would not have had surnames that matched the names they arrived in America with, would not have had surnames as opposed to patronyms or some other convention, or would not have had last names at all.
Not to mention all the flukes of family and naming conventions and legitimacy and remarriage after a death etc etc etc.
Individual history is not science, unfortunately.
posted by Sara C. at 12:14 PM on August 30, 2012
In 1622, a lot of my ancestors would not have spelled their surnames in a consistent way, would not have had surnames that matched the names they arrived in America with, would not have had surnames as opposed to patronyms or some other convention, or would not have had last names at all.
Not to mention all the flukes of family and naming conventions and legitimacy and remarriage after a death etc etc etc.
Individual history is not science, unfortunately.
posted by Sara C. at 12:14 PM on August 30, 2012
How do these databases even work? Presumably each one is only pulling from the pool of people who have submitted samples, which would probably limit it to the point that you're only getting a small non-overlapping part of the picture from each one. Even then I'd want a lot more details about how they do this stuff before making any guess about accuracy (how many and what kind of markers are tested, where is their historical data coming from, etc), a lot of this kind of testing is more about marketing than science.
posted by shelleycat at 12:16 PM on August 30, 2012
posted by shelleycat at 12:16 PM on August 30, 2012
Response by poster: Sara C: Or a totally random sampling?
This.
There may also just plain not be too many people who've submitted their own DNA to the company you're using, for them to match to
This was the point of me submitting the data to multiple companies, in hopes of finding one or a few that had some consistency between them, but so far I've hit nothing. Sure, if I had a hit that was closer than 13 generations, then I may be able to say something conclusively about last names.
What I'm hoping to find here is some sort of general metric that says "Beyond X number of generations, the chance of finding two people with the same last name is Y%".
posted by thanotopsis at 12:19 PM on August 30, 2012
This.
There may also just plain not be too many people who've submitted their own DNA to the company you're using, for them to match to
This was the point of me submitting the data to multiple companies, in hopes of finding one or a few that had some consistency between them, but so far I've hit nothing. Sure, if I had a hit that was closer than 13 generations, then I may be able to say something conclusively about last names.
What I'm hoping to find here is some sort of general metric that says "Beyond X number of generations, the chance of finding two people with the same last name is Y%".
posted by thanotopsis at 12:19 PM on August 30, 2012
Unfortunately, the closest Y-DNA relative we've isolated from the tests is 13 generations back.
Also, is this company really trying to tell you they have DNA samples from someone 13 generations back? You need to figure out where these names are supposedly coming from before you can even begin to think about what they mean.
posted by shelleycat at 12:25 PM on August 30, 2012
Also, is this company really trying to tell you they have DNA samples from someone 13 generations back? You need to figure out where these names are supposedly coming from before you can even begin to think about what they mean.
posted by shelleycat at 12:25 PM on August 30, 2012
Genetics math. 50% of my DNA gets shoved into my son when he's born.But your question is about Y DNA, which should pass down the paternal line substantially unchanged. There is not a direct proportion of Y DNA that relates exactly to a timeframe. A single marker could signify a greater or lesser amount of time. But if somebody has a proportion of all DNA in common with you, but the Y DNA is not a match, then they're not related by an unbroken paternal line. That's not to say there is illegitimacy, but rather a normal change of lastnames as women wed.
You can't say that somebody has a Y DNA match and X% of all DNA, and then use the generations indicated by the proportion of all DNA to suggest that you must share a common male forebear at that time. You may be related by a paternal line 1000 years ago, but again by a more roundabout line much later.
posted by Jehan at 12:28 PM on August 30, 2012
Response by poster: Also, is this company really trying to tell you they have DNA samples from someone 13 generations back?
No, they have a sample from someone who, based on our tested Y-DNA sample, shares a common ancestor with me 13 generations back.
posted by thanotopsis at 12:29 PM on August 30, 2012
No, they have a sample from someone who, based on our tested Y-DNA sample, shares a common ancestor with me 13 generations back.
posted by thanotopsis at 12:29 PM on August 30, 2012
Ah right, so it's a guess based on statistics and using a bunch of assumptions about heredity (with varying levels of validation). At least some of your variation is probably due to slightly different algorithms between companies, the rest is mostly from the small subset of people who have submitted to each company. As the data set grows it should become more accurate, assuming their assumptions and genetic markers are meaningful (which we have no idea about with the info you've given), but right now I'd take it with a gigantic grain of salt.
posted by shelleycat at 12:34 PM on August 30, 2012
posted by shelleycat at 12:34 PM on August 30, 2012
Another thing - your last common ancestor with these people? You probably share this ancestor with thousands, maybe tens of thousands of people. This article uses Jesus being the descendant of King David as an example... see, probably Jesus and everyone Jesus knew was a descendant of King David. Now, Jesus didn't have a last name AFAIK, but it would have been really inconvenient if everyone in Bethlehem had the same last name.
You've got a sample size problem here. Out of the thousands (if we're being conservative) of descendants of this Ancestor Guy, FIVE (including your uncle) have submitted their data to one of these companies. Once you submit your data, there will be six, and if you have the same last name as your uncle, then two names out of that six will match! Hooray!
And it's a nonrandom sample - in the sample you're looking at, there might even be an under-representation of "same names" since if one person submits their DNA their relatives might figure they'd get pretty much the same result (if you are male and looking at Y-chromosome only, there is no point in you submitting your DNA to these companies because your Y-chromosome is basically identical to your uncle's/brother's/father's). And I suspect that these services hold slightly more interest for people who don't know their family history (for instance, because no one knows who someone's father was).
So, basically: if you had the full set, it would be incredibly improbable *not* to find two people with the same last name. But you don't have the full set. You have a tiny, tiny chunk of the full set, and a weird-shaped chunk at that.
posted by mskyle at 12:36 PM on August 30, 2012 [3 favorites]
You've got a sample size problem here. Out of the thousands (if we're being conservative) of descendants of this Ancestor Guy, FIVE (including your uncle) have submitted their data to one of these companies. Once you submit your data, there will be six, and if you have the same last name as your uncle, then two names out of that six will match! Hooray!
And it's a nonrandom sample - in the sample you're looking at, there might even be an under-representation of "same names" since if one person submits their DNA their relatives might figure they'd get pretty much the same result (if you are male and looking at Y-chromosome only, there is no point in you submitting your DNA to these companies because your Y-chromosome is basically identical to your uncle's/brother's/father's). And I suspect that these services hold slightly more interest for people who don't know their family history (for instance, because no one knows who someone's father was).
So, basically: if you had the full set, it would be incredibly improbable *not* to find two people with the same last name. But you don't have the full set. You have a tiny, tiny chunk of the full set, and a weird-shaped chunk at that.
posted by mskyle at 12:36 PM on August 30, 2012 [3 favorites]
No, they have a sample from someone who, based on our tested Y-DNA sample, shares a common ancestor with me 13 generations back.They can't tell you the exact number of generations. They can only give you a ballpark figure. You may be more recently or less recently related to these people. Or maybe, not really at all.
posted by Jehan at 12:39 PM on August 30, 2012 [2 favorites]
What I'm hoping to find here is some sort of general metric that says "Beyond X number of generations, the chance of finding two people with the same last name is Y%".
Right. But you're not dealing with society's naming conventions. You're dealing with who is getting whom's Y chromosome. These two things are not the same thing. The families we recognize socially and culturally are not the same as the biology underneath them.
posted by fontophilic at 12:45 PM on August 30, 2012 [2 favorites]
Right. But you're not dealing with society's naming conventions. You're dealing with who is getting whom's Y chromosome. These two things are not the same thing. The families we recognize socially and culturally are not the same as the biology underneath them.
posted by fontophilic at 12:45 PM on August 30, 2012 [2 favorites]
Also it depends on the name and the culture - some cultures (and families) are really good at genealogy and keeping track of who's father is who, and others are a bit more casual about it. Like, if your 13-generations-ago ancestor was named Cohen, there's a good chance that he would have a lot of descendants named Cohen.
If he went by Rutting John the Whoremonger, maybe it's less likely for him to have a lot of descendants who share his name.
posted by mskyle at 1:05 PM on August 30, 2012 [2 favorites]
If he went by Rutting John the Whoremonger, maybe it's less likely for him to have a lot of descendants who share his name.
posted by mskyle at 1:05 PM on August 30, 2012 [2 favorites]
Best answer: When a person has a clear link to people in their past through last names, they probably won't seek out DNA testing for exploration. So it is possible that the majority of folks in those databases are folks without a clear name-based ancestry record. So that's a possible contributor.
posted by davejay at 1:07 PM on August 30, 2012 [2 favorites]
posted by davejay at 1:07 PM on August 30, 2012 [2 favorites]
It really depends on the company you've tested with.
I've tested 7 different Y-lines at two different companies and have gotten matches on 4 of them that I can verify with paper records. None of the other people had tested with multiple companies.
For the record, I've had the best luck with FamilyTreeDNA with three matches whereas Ancestry.com only gave me 1.
Since your Y-chromosome only gives you a small sliver of DNA to compare against, I'd highly recommend looking into autosomal testing going forward. (IMO, Y-chromosome testing is really only useful for instances where you are trying to confirm an existing paper record.) Depending on who your Uncle tested with, you could just pay an upgrade fee and not have to submit another test.
The amount of success I've had with autosomal testing is staggering. Both 23andme and FamilyTreeDNA have given me more "real" matches than I can count. (FamilyTreeDNA currently does a better job with genealogy but 23andme is rapidly improving.)
posted by fellion at 5:47 PM on August 30, 2012
I've tested 7 different Y-lines at two different companies and have gotten matches on 4 of them that I can verify with paper records. None of the other people had tested with multiple companies.
For the record, I've had the best luck with FamilyTreeDNA with three matches whereas Ancestry.com only gave me 1.
Since your Y-chromosome only gives you a small sliver of DNA to compare against, I'd highly recommend looking into autosomal testing going forward. (IMO, Y-chromosome testing is really only useful for instances where you are trying to confirm an existing paper record.) Depending on who your Uncle tested with, you could just pay an upgrade fee and not have to submit another test.
The amount of success I've had with autosomal testing is staggering. Both 23andme and FamilyTreeDNA have given me more "real" matches than I can count. (FamilyTreeDNA currently does a better job with genealogy but 23andme is rapidly improving.)
posted by fellion at 5:47 PM on August 30, 2012
Just to clarify, by real matches, I mean people that have a paper trail.
Without a paper trail it doesn't really mean anything if you had a common ancestor 10 generations ago.
posted by fellion at 5:50 PM on August 30, 2012
Without a paper trail it doesn't really mean anything if you had a common ancestor 10 generations ago.
posted by fellion at 5:50 PM on August 30, 2012
Response by poster: Well, I sent mine into Ancestry last week because their new beta program (with which I'm getting a 50% discount) is autosomal. They've been pushing the beta program for half a year, now, and I only just got in.
posted by thanotopsis at 6:13 PM on August 30, 2012
posted by thanotopsis at 6:13 PM on August 30, 2012
Once Ancestry gets out of beta, I would get your dad and someone biologically related to your grandmother but not biologically related to your grandfather tested so you can triangulate matches. (If they match you and your dad, but not this relative, you have something to work with.)
posted by fellion at 6:27 PM on August 30, 2012
posted by fellion at 6:27 PM on August 30, 2012
Best answer: A Non-paternity event happens anywhere from 0.8% to 30% of the time, depending on what population is being studied, but 3% is not an unreasonable estimate.
So assuming a 3% rate, and going back 13 generations, male-line only, means that there is a (1-.97^13)=32.7% chance that at least one of those father-son pairs might have different last names.
posted by fings at 6:50 PM on August 30, 2012 [1 favorite]
So assuming a 3% rate, and going back 13 generations, male-line only, means that there is a (1-.97^13)=32.7% chance that at least one of those father-son pairs might have different last names.
posted by fings at 6:50 PM on August 30, 2012 [1 favorite]
« Older Starting a non-profit in architecture - where to... | Feeling like a fraud when faced with compliments... Newer »
This thread is closed to new comments.
5) Birth mother filling out a birth certificate uses her own last name for the baby's last name.
6) Birth mother filling out a birth certificate uses some other name of her choosing for the baby's last name. (This one may fall under your #4?)
posted by SuperSquirrel at 11:30 AM on August 30, 2012 [3 favorites]