Thrift /Frith /Firth Surname DNA Project
Project administrator: Richard Thrift   rtx at cox dot net  (please include "DNA Project" in the Email subject line)
For the DNA Project's site map page, click here

Summary of Family Grouping based on Y-DNA marker values
using color to show differences between family groups.

The URL for this "freepages" site is long and inconvenient.
To refer to this page you can use the abbreviated form of the URL:
in place of the long version:

For each short tandem repeat (STR) marker tested, the number shown in the table embedded below is the individual's value for that marker.
Differences from the modal values for the family's haplogroup are color-coded; see the legend.
The markers that persist in a given family, differing from their haplogroup's modal values, are that family's CHARACTERISTIC MARKERS.
Characteristic markers can be used to combine the lineages of related testees into family groups and subgroups, and to separate unrelated lineages, by simple visual observation.

The DNA results tell us that:
  • Despite earlier claims to the contrary, DNA testing shows conclusively that Nathaniel Frith/Thrift of northern Virginia, Robert Thrift of Georgia, and Isham Thrift of North Carolina are NOT related to each other. Nor is Nathaniel Frith/Thrift related to any Friths yet tested.
  • However Isham Thrift is very closely related to William Thrift of Dinwiddie County, Virginia.
  • Based on paper genealogy, the descendant of Richard Thrift (born ~1828 in South Carolina, who moved to Alabama) was expected to match the descendants of Isham & William Thrift. The results do not bear this out. To better understand this situation, we need to test other Thrifts from South Carolina and Alabama.
  • Both Thomas J. Thrift of Gloucester Co, Virginia, and Absolom Thrift of Maryland were confirmed to be in the lineage of Nathaniel Frith /Thrift (as expected), despite a lack of documentation in the written records.
  • Nathaniel Frith/Thrift, born before 1650 and resident of the colony of Virginia, is a VERY close match to Thomas Thrift of Yately, Hampshire, England ~1739 (and his descendants in British Columbia, Canada). The most recent ancestor these two branches have in common MUST have lived before 1650. Due to proximity I suspect that the Yately Thrifts may prove related to the Thrifts of Kent. We need participants from the Thrifts of Kent in order to test this.
  • The descendant of Alfred Thrift (apparently, son of Edmund Rogers and Mary Evilthrift), as expected, matches no one in this project. However, neither does he match anyone currently in the Rogers DNA Project who has tested more than 12 markers.
  • The known major groups of Thrifts in the UK and the Bahamas /Monroe Co, Florida are not yet represented in this project. We need Thrifts from Kent, from Scotland, and from elsewhere to participate!

  • Madison Frith of Alabama and James Frith of Georgia (who both claimed birth in South Carolina about 1810) are clearly related to each other, despite the fact that no documentation of this relationship is known. The most recent common ancestor was before ~1810. (Kit #61429 did not actually test enough markers to establish a relationship to kit #187533 on the basis of the DNA test alone. But given the characteristic family markers, and the fact that they share the same very uncommon surname, it can be safely assumed that the two are related.)
  • John Frith of Alabama is not related to the South Carolina Frith line (including Madison Frith of Alabama) or to any other line tested. The line of John Frith & Realie Klinner from Alabama has an EXACT 67-marker match to one non-Frith surname, and a 64/67 match to another. The possible connections are being explored.
  • The Frith line from South Carolina has an EXACT 67-marker match to a family with the surname Baker, whose ancestors have been in the Garrard /Madison /Rockcastle /Lincoln county region of Kentucky, USA since 1795, and who does not match other Bakers who have tested so far. The ancestor shared by the KY Bakers and the SC Friths must be before 1810, since the agreement between the DNA results of the two SC Friths validates their paper trails to that date. At least one lineage of Friths (from Bedford Co, Virginia, USA) were known to be in this area of Kentucky since about 1800, and it is documented that during this period these Bakers & Friths lived in some of the same small communities. The most parsimonious explanation (but not necessarily the correct one) is that there was an undocumented second marriage, adoption, or similar event in the Kentucky Baker line. Undocumented events were not at all uncommon on the frontiers, when recordkeeping and record survival (not to mention survival of young parents) were hit-or-miss. This explanation of the current DNA results implies that the Friths of Bedford Co, Virginia (and of Garrard /Madison /Rockcastle /Lincoln Co., Kentucky) are of the same lineage as those of South Carolina. This is our current hypothesis; to confirm it we need Friths from Kentucky and Virginia to be tested. Please consider participating!
  • There is an unusual but recurrent given name among various thus-far-unconnected lines of Friths in the USA. Haverd Frith (born before 1795) was named in the 1812 will of Susannah Frith of Abbeville, SC. The tentative DNA connection between the SC Friths and the KY /Bedford Co, VA Friths (see above) raises the question of whether these South Carolina and Kentucky Friths were related to Harbert Frith of Bedford Co, VA (born ~1778), to Harbard Frith Sr. (born 1815-1820 in VA) who settled in Winn & Grant Parish, LA, or to John Edward Frith (born in VA ~1772). John Edward Frith's descendants, including Harbert C. Frith and Christopher Harbert Bates settled in Amite Co, MS, Avoyelles parish, LA, and elsewhere. There are also Herbert Frith and (possibly the same) Herbert Frith appearing in Bedford Co, VA marriages in 1810 and 1816, respectively. Many if not all of these families might reasonably be expected to be related. This needs to be tested. If you descend from any of these lines, please consider participating!
  • Some of the closest matches to the Friths of South Carloina are surnamed Freed. Their haplotypes are shown in subgroup A of the Freed Family DNA Project. The similarity of the two surnames raised the question of how closely these two families are related; was one an offshoot of the other? A variance analysis of the Freed & Frith haplotypes (TMRCA is related to variance, details will be posted later) suggests that the Most Recent Common Ancestor of the two families most likely lived before surnames were in use. Variance analysis is very useful when there are multiple testees available for each of two groups.
  • Friths from the known major regions in the UK (Yorkshire, Derbyshire, Lincolnshire) and its former colonies (Bermuda, Turks and Caicos, Bahamas /Monroe Co, Florida, and Virginia) are not yet represented in this project. We need Friths from these regions to participate!

  • The DNA results show that all six Firth famiies tested so far have a common ancestor, possibly in West Yorkshire long before 1767, although no two can document a common ancestor, and each decided independently to test. What are the chances of this happening? Well, that depends on how many unrelated Firth lines there are in England & the US. The finding that all six randomly selected Firth families are from a single lineage would suggest that there are few distinct Firth lines. This is consistent with the relatively tight Firth surname distribution data.
  • The lineage is an old one, having had time to acquire numerous variations. This suggests that we should be able to divide the single lineage of Firths into branches (each with their own characteristic markers). We need to test more Firths to 67 or 111 markers, in order to demonstrate this.
  • One father/son pair of Firths was tested, and a mutation was detected that arose in the son (DYS456). (It has to happen some time.) Now the son's branch of Firths has its own characteristic marker, to distinguish it from all others.
  • The Firths from Edinburgh have the most differences from the Firth modal haplotype; they form a clear subgroup by themselves. The father/son pair of testees descended from James Firth born in Edinburgh (this was claimed in a Bible entry from the mid-1800s, not confirmed yet) have a distance of 5/66 and 6/66, respectively, from the Firth mode. (That is, 5 or 6 mismatches out of 66 markers compared to the Firth modal haplotype.)(Here the son has tested more markers than the father, and I am assuming the father's haplotype is identical to the son's except for the one known mismatch.) The next most distant testee, the descendant of Nathan Firth of West Yorkshire, is only 3/66 from the Firth mode.
  • It appears that the non-Edinburgh testees can be divided into two groups based on whether they have a value of 17 or 18 at DYS458. More data is needed (not only more testees and more markers, but also better-researched pedigrees) before we can say whether this division reflects two major subgroups in reality.
  • Some of the closest nonsurname matches to the Firths are surnamed Rix from Norfolk, England. (There is a distance of 11/110 between the Rix mode and the Firth mode.) Their haplotypes are shown in the Norfolk 5 subgroup of the Rix Family Tree DNA Project. A variance analysis (TMRCA is related to variance, details will be posted later) of the Firth & Rix haplotypes suggests that the Most Recent Common Ancestor of the two families most likely lived before surnames were in use. Variance analysis is VERY useful when there are multiple testees available for each of two groups, as in this case.
  • An even closer nonsurname match (a distance of 8/110 from the Firth mode) is surnamed Kinder. He does not match other Kinders, and is almost equally distant from from the Rix & the Firth modes. We do not have enough information to speculate about the relationships.
  • Firths from Orkney are not yet represented in this project. We need your participation!

In the table below, the family-specific markers are indicated by a colored vertical column within a family group, showing similar color shades and similar marker values.
(Differences between markers that are in the "Fast" mutation class can be discounted somewhat, since we are more likely to see mutations in them. This class includes 458, 449, 456, 576, 570, CDYa/b, and 534.)

[Use the scroll bars at bottom and at right of the table]

The identification of the groups shown on this page is one of the major goals of this project. The proof of the existence of five independent Thrift lineages (two within the same US state), and the demonstration that two geographically separated Thrift families have a common ancestor before 1650, shows the power of this technique. Similar groupings could be shown for the other surnames IF more people were tested.

It is also possible to investigate, for each of these families, close matches to individuals with other surnames. Most such close matches indicate a relationship from before the establishment of surnames (which in England was circa 1300 AD). In a few cases, close matches with other surnames can provide clues about more recent events.
For more discussion of the closest matches of Firths to other surnames, see Firth Family grouping -(but this is now outdated).

An individual's collection of tested marker values is termed his haplotype. The similar-sounding term haplogroup refers to an identified group of individuals who share roughly similar haplotypes, not by accident, but because they have all descended from a common ancestor, usually in prehistoric times. (An approximation of that ancestor's haplotype can be inferred by finding the modal (most common) haplotype for the haplogroup, although this proxy ignores effects of complications such as genetic drift and sampling bias.) While a testee's haplogroup can often be predicted (guessed) based on the presence of characteristic STR marker values in his haplotype, this is not always reliable. In practice, haplogroups are confirmed by testing for a different type of marker (a single nucleotide polymorphism, or SNP) shared by everyone in the haplogroup.
Haplogroups are often referred to by groups of alternating letters and numbers, such as R1b1b1 or R1b1b2, intended to convey some idea of the relationships between groups and subgroups. However such names change often as new discoveries are made, which cause our understanding of the relationships between the subgroups to change. The haplogroup names that were in use in 2010 are in many cases outdated in 2011, and it is sometimes hard to know whether you are looking at a current haplogroup name or an outdated one. In fact since the official haplogroup-naming body is currently about 3 years out-of-date ( ), the name used for a given subclade depends on which unofficial refererence is being used as a guide. See for example . A different convention is to refer to the major haplogroup and the most specific tested SNP; thus the subclade that FTDNA last year called R1b1b2 and this year calls R1b1a2 can be referred to as R-M269, and I2b2 is I-L38. I have not yet made these web pages self-consistent as far as naming haplogroups and subclades.
Haplogroup I2b, (identified by the SNP M436 or equivalent) has at least two subclades. The subclade I2b2 is identified by the SNP L38 (or the equivalent SNP L39), however it can usually be predicted by inspection of STR values, with characteristic markers DYS455 = 10, DYS454 = 12, and YCAa/b = 19/19. Haplogroup R1b1b2 (identified by the SNP M269) can be predicted by inspection of STR values, but its subclade R1b1b2a1a2f (identified by the SNP L21) is difficult to differentiate from R1b1b2 based on STR values. The Hampshire, England & northern Virginia, USA Frith/Thrifts have been tested and found positive for L21, but the other R1b1b2 families here have not yet been tested.
SNPs are much more reliable markers for haplogroups than STRs because they mutate much more slowly, so duplicate SNPs occurring in different haplogroups are rare. On the other hand, STRs are useful in family studies precisely because they mutate so quickly (due to their repetitive structure); family branches which occur within a genealogically relevant time frame can be differentiated by STRs but generally not by SNPs. The average mutation rate for the STR markers is very roughly 1 per marker per 300 generations; if a whole set of 67 markers is monitored, this predicts roughly one change will be detected per 5 father/son transmissions, on average. In practice it can happen far more often, or far less often, than this. It is possible to find 6th cousins who have identical haplotypes, or a father and son who differ by two markers.

Closely related individuals will have almost identical STR marker values. The problem is that many 'unrelated' people in the same haplogroup will share most of the same marker values (because they do in fact share a common ancestor -it's just that the Most Recent Common Ancestor (MRCA) for the whole haplogroup was thousands of years earlier), so we have to be able to tell the difference between being closely related (that is, within "genealogical time") vs. just being in the same haplogroup.

To do this in a qualitative way, we
1) Note the modal (most common) marker values for the relevant haplogroup or subgroup;
2) Identify the markers, shared by most members of each specific family, which consistently differ from the haplogroup modal values;
3) Specifically consider these characteristic markers, which WERE inherited from the family's MRCA, when trying to decide whether others are members of this family.
4) If enough family members are tested, we can observe characteristic markers for specific branches of the family. Being able to divide a lineage into distinct branches is definitely something to aim for.
This process helps us find the most likely relationships by visualizing patterns in the data; we can crudely rank the more probable vs. less probable explanations of the limited data available. (A closely related mathematical treatment addressing off-modal markers is Ken Nordtvedt's "More Realistic TMRCA Calculations".) This qualitative approach works well for the obvious cases, not so well for the borderline cases, as in more distant non-surname matches. (In this qualitative approach we are implicitly relying on the fact that sharing a surname is an independent factor which increases the odds that two haplotypes of a given haplogroup are related. Lack of a shared surname decreases those odds, thus more care is needed when investigating non-surname matches.) The best way to make a valid interpretation for the borderline cases is to get more data, since we are dealing with statistics. Always, the way to make our interpretations of the data fit reality better is to get more data -test more people, test more markers, and extend the documentation of the family trees. And there will be cases where the real explanation is not the statistically most likely one.

[It is possible to compare the haplotypes of two people, using mutation rate estimates, to find the most likely "Time to the Most Recent Common Ancestor" (TMRCA) (in terms of either years or generations before the present), as well as the 95% confidence limits for the TMRCA. In almost all cases the 95% confidence limits are SO broad that the "most likely TMRCA" is misleading and should be ignored; rather focus attention on the stated confidence limits, that is, the earliest and latest time likely for the Most Recent Common Ancestor. This allows you to address the question, "Is it reasonable to assume that the two people compared are --or are NOT-- related in the time period of interest?" If no confidence limits are stated (as in results reported by's DNA testing /matching service), then the TMRCA prediction is useless; in these cases you WILL need to decide for yourself whether enough markers were compared to give a meaningful result.
To see examples of how broad the probability curves are, go Here, enter the Number of markers and Number of markers that match, keep the site's default mutation rate of 0.0033, click "Create graph," and notice the number of "transmission events" indicated under the leftmost and rightmost portions of the red curve. Divide the number of transmission events indicated at these points by two to get the number of generations back to the earliest and latest likely MRCA. Even with testing 67 markers, unless the matches are very close the region of reasonable probability covers a huge range of generations.
Characteristic markers for descendants of:
Thrift of Hampshire, England, and Frith/Thrift of northern Virginia, USA: DYS439=11, DYS448=18, DYS460=10, YCAIIa=17, CDYa=38, DYS413b=24. DYS576=17 is a characteristic marker for the Virginia, USA branch, but may not be characteristic for the British Columbia, Canada /Hampshire, England branch.
Thrift, presumed of southern Virginia, USA: DYS390=23, DYS385a/b=14/15, DYS439=13, DYS458=19-20, DYS456=14, DYS570=18, CDYb=37, DYS438=13, DYS450=14, DYS520=19, DYS568=12.
Frith of South Carolina, USA: DYS390=25, DYS385b=15 [others would be identifiable if we had another 67-marker haplotype].
Robert T. Thrift of Georgia, USA: DYS390=23, DYS385b=13, DYS426=11, DYS388=15 [others would be identifiable if we had another 67-marker haplotype].
Firth of West Yorkshire, England: DYS393 = 14, DYS19=15, DYS458=18/19, DYS459b=9, DYS447=25, DYS449=29, Y-GATA-H4=8, DYS607=13, DYS576=15/16, CDYa=35, DYS557=14, and DYS617=14. [Some of these may be specific to the branches represented by the current testees.]
NOTE -These characteristic marker values use the current FTDNA convention. For comparisons against or other databases, values for four specific markers need to be converted to format. To the FTDNA value of DYS441 add 1; to DYS442 add 5; to Y-GATA-A10 add 2; and to Y-GATA-H4 add 1.

I'm trying out a new format for showing an Excel table on a web page. I hope it shows up well on your particular combination of browser & monitor.
Please let me know of any problems: Richard Thrift, rtx at cox dot net (please include "DNA Project" in the Email subject line).
(It may be difficult to read the STR marker names in the table, because of the small font used to allow more columns to be visible at once. Usually the marker names don't need to be legible; when desired you can alter the zoom setting of your browser.)