A civil discourse on IrishOrigenes' methods

[Dave-V]

Dr. Tyrone Bowes has been posting his IrishOrigenes case studies on Facebook, and I seem to be the only one there who thinks the emperor has no clothes. His methods are not peer-reviewed, so I wanted to check my concerns with the experts on this forum. Am I off-base in thinking this doesn't really work?

On the plus side, I DO think these studies are a creative approach to combining DNA and surname analysis. I think his approach is beautifully simple and it would be great if it was correct and certainly great if it encourages more people to test their DNA. And I think it’s is certainly applicable in limited cases where you already have other evidence that the theory will hold.

Ok… sigh. Here are my concerns. I picked a lot here on the Mackenzie study, but the others are similar.

1. I can’t grasp how matches of other surnames point to a “genetic homeland” back when surnames were first adopted.

For example my closest match of a different surname is a Whalen at 3 GD and about 12 generations away, so the simplest explanation is that we had a common ancestor around 1600AD and at some time later either a Vance on his line became Whalen or a Whalen on my line became Vance. Of course that’s not the only explanation. And the surname swap could easily have happened down one of our lines much later than our common ancestor’s lifetime.

But I get how we might say that my Vances and his Whalens were living near each other when the swap happened at that unknown point in history after 1600AD, although that leap really isn’t obvious (people did travel). But then somehow all of this suggests that Vances and Whalens also lived near each other in a “genetic homeland” back in 800-1000AD too?? Did I miss a meeting?

I get the idea that native Irish and Scots were tied to their land and statistically more likely to have stayed there. Of course, Irish and Scottish farmers are also statistically the most likely to have had their surnames derived from the ancient Gaelic for “those darned aliens are out making crop circles in my fields again”. But you’d need evidence to say any of them actually did.

Another key idea is that recent censuses reveal patterns back to when surnames were first adopted. So then shouldn’t every surviving list of people in Ireland and Scotland back through time consistently show the same patterns? Do they? Has anyone looked?

And if recent Irish and Scottish surname distributions reflect the same patterns as a thousand years ago it would mean that none of the migrations into and out of Ireland and Scotland over a millennium, none of the invasions, uprisings, and wars, none of the plagues, famines, and emigrations, none of the plantations, dispossessions of land, and forced relocations – none of it had any effect on the distribution of native surnames around Ireland and Scotland. And everyone’s ok with that?

Just one son moving to a different county could put 30 or more families of that same surname in the new county after just 4 generations, and there are ten times that many generations back to when surnames were first adopted. Really – nothing changed over that time period?

2. The studies rightly cite that anyone’s Y-DNA line back to the start of surnames has about a 50% chance of an NPE, but they don’t acknowledge how much NPEs might throw off the results that apparently DO fit the pattern as well. Doesn’t the same point also mean that 50% of the matches used to find “genetic homelands” are probably NPEs as well? Or what if some of the same surnames came from completely different origins? How would anyone know?

3. Why does the analysis pick from different DNA marker tests? For instance the MacKenzie study cites a Prendergast match at 37 markers although it doesn’t appear at all at 67 markers. And why are 37 or 67 better than 111 markers? Wouldn’t different markers mean different matches and therefore a completely different “genetic homeland”?

4. Why not compare SNPs on the matches that pinpoint the “homeland”? What if they don’t match on all known SNPs over 1200 years old? Wouldn’t that be a good test of the idea that these matches really did have a common ancestor when surnames were adopted?

5. The case studies assume that if you match anyone of a surname, you automatically match everyone of that surname and all its presumed variants. Again back to the MacKenzie study, which concludes from one close match to a McCloughlan that the test subject and entire Clan MacKenzie and Clan MacLachlan have common ancestors. And all without any consultation of the relevant surname DNA projects, because clearly all those poor project admins are wasting their time with their participants’ various paper trail and DNA origins. Just get a helpful map and look up the surnames!

6. It just feels like case studies wander through genetics, unsourced statements of fact, and family legends looking for a possible explanation. Here for example is my walkthrough of the MacKenzie study, exaggerating the leaps to make the point:

• Let’s start by comparing the one MacKenzie test subject to his reoccurring surname matches, treating all the 37 and 67 matches and everything from 1 GD to 7 GD as the same thing. No reason to confuse things with facts like all their different common ancestors probably lived centuries apart.
• A McCloughan is his closest match, which clearly represents the whole Scottish Clan MacLachlan.
• He matches a lot of other names that are Norman, English, Scottish, Irish, Welsh, or some other origin. So we’re obviously dealing with Normans here.
• Did you know that lots of Normans lived in Co. Wexford? Here’s an exclusive unexplained color map of Wexford that applies to every single person who has these surnames. Oh look, there are Fitzgeralds on here, too.
• Hmm. It just happens that Clan Mackenzie has a legend of descent from the Norman-Irish Fitzgeralds. Let’s follow that legend over to Scotland.
• Clan Mackenzie lived in north Scotland. Clan MacLachlan lived waaaay further south. We’re now one complete island and four pages removed from any genetic analysis, but we’ll still call these people “Mr MacKenzie’s genetic relatives”.
• Aha. We’ll pull out the 1841 census of Scotland where there are Mackenzies and MacLachlans [along with about 40,000 other people on average based on Scotland’s 1841 population density] in the middle of Scotland nowhere near either Clan. Clearly they’re all related to each other and to our test subject. Bingo!

I’m not criticizing the case study on clarity or structure. Frankly with all the leaps and twists I feel like I’m reading Dan Brown’s Da Vinci Code. So shouldn’t all the other surname matches besides McCloughan/MacLachlan be there in the “genetic homeland” too, if the theory is correct? Oh but wait - if these families came from the Norman-Irish Fitzgeralds, shouldn’t we be looking at their origin for the “genetic homeland” if it’s where surnames first originated? What methodology was this again?

7. As to the unsourced statements of fact - hey, I’m not sourcing the statements of fact I make here either; I’m ok with letting some go. But not multiple statements that are central to the analysis. Some examples just from the MacKenzie and Foy studies that I think are suspect but at least need supporting evidence:

“In Ireland each of the estimated 1,500 distinct Clans had a single founding ancestor”

“one must first identify the surnames that continually appear as genetic matches. These reoccurring surnames are less likely to be a result of non-paternal events”

“Almost everyone with Scottish ancestry will be genetically related to at least one of these prominent Clans and families.”

“non-paternal events are more likely to occur between neighbouring Clans”

There are a lot more of these, but you get the idea.

8. The case studies open with the claim that “at present there is an 80% success rate”. What is “success”?? None of these “genetic homelands” are verified, this is all supposition. Doesn’t that just mean “80% of whatever case studies were included in this calculation reached an unverified conclusion”??

The case studies are very clear that the methodology doesn’t work in every case. But that also means the results you like can’t prove the theory. You have to be able to predict when the theory will work and when it won’t before you test, and then show that the results bear that out. Otherwise it’s not science.

9. Each case study ends with a statement like “your genetic homeland can be verified by recruiting local male volunteers for DNA testing”. Well, first maybe the studies should also be clearer that they’re really not verified. And actually local testing won’t prove it; you’d also have to test more widely to prove that that was the ONLY concentration of the same DNA and it’s the real (or only) “genetic homeland”, but still that would certainly be independent evidence. Has this ever been done?

Ok, I’ll stop there. In my opinion, this is an interesting methodology based on a dubious and unproven theory that has identified a whole bunch of possible “genetic homelands” without even one actual case shown to be correct. Is it possible that some of the case studies are correct? Sure. How would you ever know?

Dr. Bowes, if I have unfairly characterized your methodology, I offer my apologies. I wish you luck and I would be happy to be wrong; it certainly WOULD be really cool if your theory is further proven. But in my opinion your case studies are just offering up new legends for our surname origins. I already have enough of those.

Dave Vance

[AJL]

“In Ireland each of the estimated 1,500 distinct Clans had a single founding ancestor”

I am very suspicious of this. Multiple surnames believed to be of a single clan have multiple progenitors that probably have nothing to do with NPEs. The excellent work the Donalds have done with DNA shows this.

“one must first identify the surnames that continually appear as genetic matches. These reoccurring surnames are less likely to be a result of non-paternal events”

This seems fair.

“Almost everyone with Scottish ancestry will be genetically related to at least one of these prominent Clans and families.”

"Genetically related" to what degree? This seems like a chicken-and-egg case. Are various people genetic relations to the clans, or are clan members related to genetic groups?

“non-paternal events are more likely to occur between neighbouring Clans”

I would say this puts undue emphasis on an idealized view of historical clan structures that is probably not of much relevance. NPEs are more likely between neighbouring males -- but it's not as if, in historical times in Celtic countries, all neighbours were of a single clan that was itself of a single haplogroup. It is true that NPEs are more likely to occur between males in proximity, but this really has little to do with clans.

There is something at the root of what he says: given sufficient men testing with good paper trails, you ought to be able to pinpoint a rough locus of origin for motifs.

However, in the case of Scotland there were so many changed surnames in the 17th-18th centuries first because of proscription, for instance, as explained here; and second, as Highlanders moved to the cities and as English became predominant even in the Highlands, people often changed their names so that clerks could actually write them down in English. Some Scots names predominant in registers of the earlier part of this period (as this one) more or less disappear by the late 18th century, as all members of that family in one locale change their names.

Incidentally I descend from the line of kit 61784 in this project.

As you can see, there are multiple Gillespie lineages and mine could not be construed as having anything to do with the most prevalent groups. Though an NPE is certainly possible this also does not necessarily imply an NPE, since change of surnames was not rare. Neither is there any one clan in common among the closest matches to 61784 that can explain things.

Something often ignored is that women kept their maiden names even after marriage, and if they were widowed, their sons often had their names rather than the father's. Also, in parts of the Highlands, patronymics continued to be used as did occupational names until fairly recently.

[Ian B]

Another answer to like surnames with different yDNA is the fact that people often took on the name of a powerful clan or family, in the case of the Irish the enforced Anglicisation of family names by the enforcement of the 1698 Penal Laws resulted in families with different families names suddenly sharing an anglicised version.

[rivergirl]

I think Irish Origenes is a simple tool that may work for some, but not for everyone.
Most could do the same research themselves with the Irish Census, Griffiths Valuation, and research into the surnames and placenames.
Wether it put them within 5km of the "genetic homeland" is a different matter.
I do agree you need to test men from all areas where the surname is found in the homeland.


In the case of your match with Whalen, you only match with 1 man from the 80 men in the Whalen/Whelan/Phelan group. He matches many Vances.
Whilst your Whalen match's family have probably been in Tipperary for many many generations, his yDNA line does not appear to be from that area. (R-L513 appears more in men from the north of Ireland)

I noticed that the Irish Origenes website have Whelehan as a variant of Whelan, yet it not considered a variant and is genetically different from the Whelan/Whalen/Phelans. I dont know where he got that information.

Whelehan: fairly numerous: S Midlands etc. Ir. Ó Faoileacháin, faoileach, joyful. A name associated with W Meath.
SGG; Ó Droighneáin, M. & Ó Murchú, M.A., An Sloinnteoir Gaeilge & an tAinmneoir, Baile Atha Cliath, 1991

Phelan: Very numerous: all areas, especially South East. Ir. Ó Faoláin, faol, a wolf. They were chiefs of the Déise (Waterford). In 1169 A.D. the first man to fall in battle resisting the Invasion.
In W Ulster, a bardic family, Ó Fialáin. SI & IF.
SI: MacLysaght, Edward, Surnames of Ireland, Dublin, 1985
IF: MacLysaght, Edward, Irish Families, Dublin, 1980

[DebbieK]

The methodology was actually discussed in the Surname Distribution Group on Facebook:

https://www.facebook.com/groups/467543026597487/

You'll need to scroll quite some way down the page. The relevant discussion was started by Richard Thrift on 24th April 2013. Although Tyrone Bowes contributed to the discussion we did not get answers to the questions we raised.

I have a number of concerns with the methodology apart from the lack of peer review.

The FTDNA database is heavily biased towards Americans of British and Irish origin so the matches that people are getting are not necessarily representative of the origins of the surname. Most surnames have multiple origins. Single-origin surnames all have multiple genetic lineages and without detailed genealogical research it is impossible to tell where the NPEs have occurred. Americans are more likely to match each other. Surnames have only had 400 years to develop NPEs in America compared with 900 to 1000 years in Britain and Ireland. Also the surviving American lineages tend to be very prolific, often having hundreds of living descendants. Americans matching Americans with other surnames is telling you more about NPEs in America than surname origins in Britain and Ireland.

The methodology for handling NPEs is not clearly defined. If a surname only occurs once in a list of matches it is discarded for no apparent reason. It seems that surnames are cherry-picked at random from match lists to fit the stories.

No account seems to have been taken of the SNP testing status of the people in the match list. I'm now seeing people who match on 34/37 markers who are in different subclades of R1b. Similar problems have been reported with more distant matches at 67 markers.

Some people have tried to reproduce the methodology with English surnames and it clearly doesn't work. As someone else said, if a method works it works. Why should success only be claimed for Irish surnames?

In order to determine the origin or origins of a surname there is no substitute for doing the detailed genealogical research on all the different lineages in combination with a structured DNA testing programme. The researcher also needs to look at the distribution of a surname at different points in time.

Match lists can sometimes provide useful clues, but inferences are being made which are not supported by the evidence. Some of those matches might be relevant and surnames do tend to cluster in particular locations but without doing the genealogical research I do not see how you can determine which matches are relevant and which ones are NPEs or have incorrect origins because of faulty research.

[Jean M]

Good to see you here Debbie.

[Bowe(s) ONS]

Here's a direct link to the facebook discussion: https://www.facebook.com/groups/4675...9417391076716/

[George Chandler]

David,

You have a lot of different questions within your post. Your degree of certainty regarding a closer match with another surname will depend on a lot of different things.

1. The number of different markers tested and being compared.
2. The mutation rate for the markers which the 2 surnames differ.
3. The pattern of the markers tested (unique values).
4. The level of SNP testing that both surnames have tested for.
5. The level, quality and type of genealogy research.
6. Oral family history
7. Geographical distrubution and concentration of SNP's and marker patterns.
8. Geographic isolation
9. Religious history and isolation
10. Geographic features
11. Surname spelling variations
12. Genetic triangulation

There are more clues used but those are a few examples of what people are looking at. Not everything is a NPE when it comes to different surname histories obviously. In some of the old documents if a person was called "William of London" for example. Some descendant his let's say became a Miller who would then be called so and so "the Miller of London" which eventually evolved into the last name of Miller. You can in theory have 2 surnames one being London and the other being Miller which have a common ancestor and yet there was no NPE. Something else to consider is the people did take aliases even way back to the begining of surnames as well.

George

[Mikewww]

On the plus side, I DO think these studies are a creative approach to combining DNA and surname analysis

Yes, any thoughtful analysis is good as it presents possibilities and encourages discussion and further research. That doesn't mean the analysis is worth paying for, depending on one's budget.

1. I can’t grasp how matches of other surnames point to a “genetic homeland” back when surnames were first adopted.

I agree. The analysis of surnames is useful information but it is not conclusive as far as a indicating a homeland.

2. The studies rightly cite that anyone’s Y-DNA line back to the start of surnames has about a 50% chance of an NPE, but they don’t acknowledge how much NPEs might throw off the results that apparently DO fit the pattern as well. Doesn’t the same point also mean that 50% of the matches used to find “genetic homelands” are probably NPEs as well? Or what if some of the same surnames came from completely different origins? How would anyone know

I think that ISOGG recommends usage of a 4% per generation where we think the father is confident he is the father. An NPE could happen in the first generation of the progenitor so there are certainly are no guarantees. This is where triangulation of different lineages helps get you into the ballpark of the progenitor, but its really only the ballpark.

4. Why not compare SNPs on the matches that pinpoint the “homeland”? What if they don’t match on all known SNPs over 1200 years old? Wouldn’t that be a good test of the idea that these matches really did have a common ancestor when surnames were adopted?

Of course, SNPs need to be a part of analysis. They are generally much, more more stable and reliable as far as indicating the appropriate branch/lineage. They may not be granular enough, but that is changing as more and more SNPs are being discovered.

7. As to the unsourced statements of fact - hey, I’m not sourcing the statements of fact I make here either; I’m ok with letting some go. But not multiple statements that are central to the analysis.

I'm not sure I understand the particulars, but I agree that all essential statements of facts should be sourced with solid backup.

8. The case studies open with the claim that “at present there is an 80% success rate”. What is “success”??

I don't know about the particulars of this, but I am familiar with people stating they have SNP prediction rates or probability rates but if they are not founded in statistical probability theory with error ranges and the whole bit, the probability statements are meaningless.

I don't know the details of the whole discussion, but I'm glad you are asking questions to better understand the value of these kinds of things and how we can improve our analyses.

[Bowe(s) ONS]

Originally Posted by Dave-V
“one must first identify the surnames that continually appear as genetic matches. These reoccurring surnames are less likely to be a result of non-paternal events”

AJL:

"This seems fair."

... I think this becomes another hit or miss part of his approach. He doesn't set any parameters for "recurring" and set forth reasons why the chosen parameters make his method stronger. His Flanagan case study is based, as usual, on only 37 markers and shows that the tester matches only two individuals each for two matching surnames at that level, Reaney and Fitzgerald (http://www.keepandshare.com/doc/4441...f-1-3-meg?da=y). With those limitations at the start he proceeds anyway. In others of his case studies, matching surnames with an recurrence that low aren't even considered when other matching surnames have a higher number of recurrences. His Kelly case study is an example (http://www.keepandshare.com/doc/6329...f-1-4-meg?da=y). With Kelly, there are matches to many other surnames and he appears to systematically exclude all those with only two representatives. It seems to me he's choosing which matching surnames to include in the mapping analysis based on the relative numbers within each study of the matching surnames. This strikes me as showing inconsistent handling of what constitutes meaningful "recurring" matches that can be deemed relevant to finding one's Y-DNA genetic origin precisely at the time one's surname was first assumed. Additionally, the potential pool of matches from each matching surname will be affected by other variables. This, from the standpoint of methodology, makes it hard to simply conclude that the surnames that recur the most do so because they represent one's ancestor's neighbors within his claimed 5 km radius when surnames were taken, as opposed to them recurring the most because the pools of possible matches are "lopsided" in various ways:

• Some surname DNA projects - whether their surnames are rare, common or in-between - are funded which allows easier recruitment because participants don't have to pay.
• Common surnames also often enjoy easier recruitment because documentary brick walls occur, in addition to the usual reasons, due to the same forename/surname combination being common.
• The no. of each matching surname at any marker level will change as the database grows, thus a matching surname that at the time of one of Tyrone's case studies is dominant among the matches, later may not be.
• Even under ideal conditions, it's not likely a surname DNA study ever has a representative sample for the surname.
• Others?

What do the matches look like in 5, 10, 20 years as the database grows? Do the same surnames recur at the highest frequencies? No doubt in some cases they will, but where does the margin begin outside of which recurring matches won't stand the test of time, if it were even possible to determine such a thing?

(Not an expert in any of this and always open to correction/education.)