The Descendants Of Somerled: DNA Evidence

The genealogy of the present Clan Chiefs is well known. It is presented in Peerage books such as the current Burke's [Burke's 2001], and on the Clan Donald Edinburgh web page. Sellar [Sellar 2000] discusses the details of the first few generations. The genealogy of large numbers of lesser lines is copied onto that web site from the 1904 book on the Clan Donald by the Revs. A. and A. MacDonald [MacDonald 1904]. In addition we have received well to fairly well documented lines all the way to Somerled from 6 other of our participants, one being Allan Douglas MacDonald, Chieftain of Vallay. These are presented in the following descendant chart. Names of near ancestors of living participants who are not clan Chiefs have been removed for privacy reasons, but we have them in our records. The three Chiefs are color coded blue, the two Chieftains, Allan MacDonald of Vallay and David Macdonald of Castle Camus, are coded green.

Somerled Descendant Chart

The following chart shows the same people, in the same order, showing the mutations we have measured for them in the Family Tree DNA set of 37 markers. We have included one marker, DYS557, from the 67 marker Family Tree DNA panel. The mutation in this marker occurred early in the Sleat line, first occurring in Donald of Harlaw, Donald Gallach 3rd of Sleat, or one of the two men in between them. Though as more participants are added to our project the modal value of CDYb has from time to time changed, the values for participants with paper trails makes it clear that the original value was probably 38. We have also included in this chart the branchpoint to the MacAllister line. The positioning of the markers along branches is chosen to minimize the number of mutation events. Computer calculations using more sophisticated methods which take into account marker rates give the same results.

Somerled lines with marker changes chart

The key person in our discussion is John, Lord of the Isles, who died in 1386. He is often known as "Good John". He was progenitor of most but not all of those Somerled descendants we have in our study. An additional "Network" chart showing the relationships for most R1a participants is available here as a popup. The calculated haplotype of Good John is the same as participant &PGTBN to 37 markers. We have descendants of three of John's sons. Nine of these have excellent paper trail pedigrees. As shown above on the two charts, there are 22 single-step mutations in 37 markers in 139 transmission events. This is a mutation rate of 0.0043±.0009 (one standard deviation) per marker per birth. The value calculated by the Webmaster from all available data, both academic papers and surname studies, is 0.0031, only slightly more than one standard deviation off from our Clan Donald value. This is well within the expected range. Two other statistical tests are possible to perform on our data. One test compares the number of mutations observed for each marker in our study to the known mutation rates; these should be compatible, and are. The number of mutations in each of our nine lines is actually expected to be different: it should follow a Poisson distribution, and it does, to a remarkable degree of accuracy.

All this means that our DNA data is fully compatible with our paper trails. The DNA offers no reason to suspect that the genealogy is different from the paper trail. It verifies that all nine participants are either descendants of the person who occupies Good John's position in our charts, or of his close relatives. In that sense it verifies the paper trail. What it really does is verify that the three people we show as sons of Good John are very closely related. The data so far could be compatible, for example, with some other man fathering these three sons. This is highly unlikely based on history, because the three sons are supposed to be by two different wives.

Our project has one R1a McAlister, one R1a MacAllister and two (identical to 37 markers) Alexanders. Alexander is considered a variant of MacAlister. The haplotypes of five more R1a MacAllisters, three Alexanders, an Allison, the latter two being common replacements for "MacAllister", and a MacDougall are either publicly available on Ysearch or SMGF or have been obtained from other DNA projects. For none of these do we have full paper trails, though the ones with listed lines have family tradition that they originate from the listed line. The data for these people is shown below. Only markers for which at least one person has a difference (highlighted in pink) from the "Somerled" haplotype are shown. The "Other" column shows how many more markers match the Somerled profile exactly.

  3
9
0
1
9
3
9
1
3
8
5
|
a
3
8
9
|
1
3
8
9
|
2
4
5
8
4
5
9
a
4
4
7
4
4
9
4
6
4
b
/
c
G
A
T
A
H
4
Y
C
A
I
I
B
4
5
6
6
0
7
5
7
0
C
D
Y
a
4
4
2
4
3
8
4
4
4
4
4
6
4
5
2
G
A
T
A
A
1
0
G
A
T
A
C
4
O
t
h
e
r
MacAllister 25 17 10 10 13 30 16 9 23 31 16 11 23 16 18 19 36 13 10           7
McAlister 25 15 11 11 14 31 15 8 23 31 12 15 21 17 16 18 34 12 12           18
McAllister 25 15 11 11 14 31 15 8 23 31   12 21 17 16 18 34 12 11 13 12 31 13 23 26
McAllister 25 15 10 11 13 30 15 8 23 31   12 21 17 16 18 34 12 11 13 12 31 13 23 26
McAlister 25 15 11 11 14 31                                     7
McAllister 24 15 11 11 14 32                                     7
McAllister 24 15 11 11 14 31                                     7
Allison (MacAllister of Loup line) 25 15 11 11 14 31 16 8 23 32 12 12 21 17 16 18 34 12 11           18
Alexander (of Menstrie, from Earl of Stirling or Caledon) 25 15 11 11 14 31 15 8 23 30 15                           27
Alexander (of Menstrie, from Earl of Stirling or Caledon) 25 15 11 11 14 31 15 8 23 31 15                           27
Alexander 25 15 10 11 14 31 15 8 23 31   12 21 17 16 18 34 11 12 14 13 31 11 23 26
Alexander 25 15 11 11 14 31 15 8 23 31 15 12 21 17 16 18 34 11 12           18
Alexander 25 15 11 11 14 31 15 8 23 31 15 12 21 17 16 18 34 11 12           18
MacDougall 25 16 10 11 13 30 15 9 24 31   11 23 16 16 18 34 12 11 12 12 31 13 25 26

The same tests described above can be applied to this data. This analysis confirms that the main R1a Clan Donald line is indeed the line of Somerled's grandson "Donald the eponymous". The marker DYS458 appears, based on this limited data, to distinguish MacDonald from MacAllister. One R1a participant from Glencoe (whose line branches off from a brother of Lord John) also shares 15 at DYS458. We now beleive that we have sufficient data to show that the ancestral state was 15 at DYS458 and that the actual mutation occured with the birth of Lord John himself, his father, or grandfather. While this is not absolute proof, we consider it reliable at the 90% or better level. DYS458 is a fairly fast marker, and we do see what are likely back and parallel mutations. The number we see is quite accurately in accord with the known mutation rate. The pattern of 11 at DYS442 and 12 at DYS438 is indicative of the name Alexander, though we do not know at all if this is a universal diagnostic.

The marker YCAIIb is extremely important for arguments involving Somerled descendants. A 23 at this marker is the norm for most R1a worldwide; the Somerled value of 21 is characteristic of Norse vikings. The fact that almost all available R1a haplotypes of surnames MacDonald, MacAllister, and variants, have YCAIIb = 21 while only 50% of all R1a haplotypes in Scotland and 30% of all R1a haplotypes on the Ysearch database (which is heavily oriented to western Europe) have values less than 23, is good evidence that most if not all of our R1a Clan Donald men are related in historic times. That only one such surnamed man had YCAIIb = 23 is a very strong statistical indication that they are not anywhere near a random subset of all R1a men in Scotland.

Proof of the line of Somerled himself depends on data on MacDougall, which arises from a different son of Somerled. The MacDougall above, while R1a, is not at all a match, rather, it is clearly quite different, especially at YCAIIb, where he is the non-MacDonald 23. Good data on many MacDougalls would be necessary for a full proof of the haplotype of Somerled, and would resolve the DYS458 ambiguity mentioned above. Sykes in his book "Adam's Curse" [Sykes 2003] mentions the MacDonald/MacAllister to MacDougall connection. He has given us his data, but even added to the very tiny number of R1a MacDougalls we have obtained elsewhere the numbers of R1a MacDougalls are swamped by R1b MacDougalls. This is different than for MacDonald itself (and MacAllister) where the number of R1a people is substantial compared to those who are R1b. The actual DNA proof of the line to Somerled himself thus remains just beyond reach.

<Previous: Before Somerled
Next: Other Ancestry >