Geographic
Patterns of Haplogroup R1b in the British Isles
Kevin D. Campbell
Abstract
The recent availability of Y-STR databases has provided the
opportunity to further explore geographic and subclade patterns of Haplogroup R1b
in the British
Isles. However, works based upon this data leave
gaps in the published analyses that make it difficult to link conclusions to
the supporting data. This paper
identifies haplotypes in the Oxford Genetic Atlas Project data and
links these to Bryan Sykes’ book, Blood of the Isles. The analysis provides support for Sykes’
conclusions and posits the genetic signatures for the Dal Riada Celts and Picts
identified in Sykes’ book.
Address
for correspondence: Campbell@alum.mit.edu
Received: January 1, 2007; accepted:
April 27,
2007.
Introduction
While DNA testing has evolved rapidly, there is a dearth
of reliable Y-STR DNA data for serious analysis. This absence is particularly surprising given
the heavy concentration of the R1b haplogroup in Western Europe and the bias of certain key databases
such as YSearch, SMGF, and FTDNA toward participants
from North
America
and Western
Europe.
The lack of reliable data can be attributed to a number of
causes. These include: inconsistent use
of markers and nomenclature, cost involved with extensive panels of markers,
and a number of other issues that are familiar to most academic and amateur
genetic researchers. However, it is
suggested that there are two root causes that have significantly hindered
population analysis – (1) the lack of uniformly collected, independently
verified data sets and (2) the tendency of researchers to shield and obfuscate
their analysis.
The major factor contributing to the first is the nature
of the submission process at the largest databases, which results in
non-validated and unsubstantiated data.
The sad truth is that many of this field’s largest databases rely on
data and geographic inputs provided by enthusiastic, but uninformed
individuals. While transcription and
upload errors for YSearch and SMGF may be very low, many errors creep in
related to marker translation and geographic speculation. Though some might dismiss marker translation
as insignificant, errors in geographic data are not. In fact the essence of population analysis is
the deduction and inferences that can be made about how the Y-STR data relates
to specific geographic areas or population migrations.
As an example, in the Campbell project 73% of the participants
(123 out of 169) do not have a reliable paper trail that places their genetic
ancestor in either Scotland or Ireland. Of those that do, 39% have a
paper trail or some level of confidence that their ancestors came from Ireland and 61% believe that their
ancestors came from Scotland.
Given that the majority of Campbell participants can only reliably
document the birth location of their oldest proven ancestor to the 18th
and 19th centuries,
it is also likely that even much of the Ireland/Scotland split is further
confused by the Scotch-Irish migrations of that period. In short, it is suggested that much of the
location data provided by DNA participants to the large voluntary databases is,
at best, of marginal reliability when performing large population genetic
studies.
The aforementioned statistics call into question the
reliability of information provided by individual participants and suggests
that possible organized and structured data gathering studies might provide
better sources of data. However, some
respected researchers have published popular texts with new theories while
providing insufficient information or omitting critical linkages that might
facilitate a formal and critical review.
For example, though Capelli’s study of Britain has been published for several
years (Capelli et al, 2003), only recently has the underlying data been made
available. While EthnoAncestry
has introduced a DNA test for “Pictishness” they have not yet published any
details regarding their analyses.
Though Bryan Sykes has made the data available that he
used in his recent popular book “Blood of the Isles”, he also leaves out
critical linkages between the particular haplotypes and conclusions that he
draws in the text. This is particularly
frustrating since while the haplogroups in his study are reasonably distinguishable,
the assertions and theories that he makes about specific subclades are not
easily reviewed in the context of the supporting data. The present article will fill in some of this
missing information.
Similarly, most people would consider the recent book, The
Origins of the British, by Stephen Oppenheimer’s (2006) to be an
authoritative work in this rapidly evolving field. While a substantial portion of Oppenheimer’s
Y-STR study also uses the Capelli data, he assigns new haplotype labels to the
data – “16 distinct types of R1b” --
without ever providing the detail necessary to link his haplotypes to the
underlying data. As is the case with
Sykes’s work, this makes any real academic or juried review of his conclusions
impossible and lessens the usefulness of his work to other researchers. In some cases like the aforementioned Pict
test, researchers have been quick to partner with testing companies to make
money from their private theories.
The purpose of this paper is to take a look at the
underlying Sykes R1b data and see if it can be linked to his founder haplotypes
and the conclusions of his analysis. The
goal of this paper is to attempt to provide additional insights to the work of
these researchers to make it more useful to the individual genetic genealogist
who look to their data as a link to the past.
Methods
This study focuses on Haplogroup R1b, which comprises the
vast majority of the British Isles. R1b data
was extracted from the targeted data set and two types on analysis were done to
identify patterns – affinity analysis and network analysis. Network analysis was performed using the
Fluxus Engineering’s phylogenetic network analysis software (Network
4.201). For the affinity analysis of
haplotypes, an Excel spreadsheet was developed to look at patterns and
anomalies of the data. An algorithm was
developed that took as input the 10 marker values of the haplotype signature
and then compared this against the overall distribution of that haplotype in
the full database. A sequential summary
of the method is as follows:
- Code
Sykes’ Oxford Genetic Atlas Project (OGAP) data by haplogroup.
- Separate
the data into haplogroups and limit the analysis to R1b only.
- Group
similar R1b haplotypes.
- Look
for geographic patterns or anomalies in the distribution of major
individual haplotypes.
- Review
the data “in the large” to see if individual patterns can be explained in
a larger context and against Sykes’ published conclusions.
Results
Step 1 -
Coding of the OGAP Data
For data collection, Oxford Genetic Atlas Project (OGAP)
data was downloaded from Bran Sykes web site and converted from PDF to
Microsoft Excel format. 2,322 samples
were then coded by haplogroup using Whit Athey’s improved Bayesian haplogroup
calculator (Athey 2005, 2006). Hereafter,
“OGAP” will refer to the R1b subset of Sykes’s data.
Sykes included the following description of the data in
the supplementary data file:
Y-chromosome DNA (yDNA) - Samples collected early in OGAP were amplified across the
following seven markers: DYS 19, DYS 389i, DYS389ii, DYS 390, DYS 391, DYS 392,
DYS 393 using conditions described by de Knijff et al (International Journal of
Legal Medicine, 110: 134-140, 1997). Later samples were typed for these and
three additional markers: DYS 388, DYS 425 and DYS 426, using the two-stage
multiplex conditions described by Thomas et al. (Human Genetics, 105: 577-581, 1999).
Alleles are reported as the number of repeat units. For reasons of continuity
within OGAP, DYS 389i is reported as three repeats lower than the allele size
produced by the ABI 3100. DYS 398ii-i reports the difference between 389i and
389ii, the reason being that the repeat size at 389ii is not independent of
389i whereas the difference between them is. Although Y-chromosomes were
assigned to clades, largely by RFLP [Author - restriction fragment
length polymorphism] analysis, these assignments are
not reported here as they do not necessarily correspond to the SNP-based system
recommended by the Chromosome Consortium (Genome Research 12:339-348, 2002).
Geographical distribution - Genetic data are assigned to geographical regions based on the
birthplace/residence of the paternal grandfather. This was done to minimize the
effect of very recent migration. The regional boundaries are shown on a map which
precedes the Prologue in Blood of the Isles.
These data are copyrighted and must not be reproduced without
permission. Other formats and additional details may be available for academic
collaborations.
Several things are worth noting about the data. First, the Sykes data only uses 10 markers
(DYS19, DYS389i, DYS389ii, DYS390, DYS391, DYS392, DYS393, DYS388, DYS425 and
DYS426). In addition, only approximately
64% of the data are complete with 36% of the data missing four markers, DYS439,
DYS388, DYS425 and DYS426. While the
missing markers appear to be a serious shortcoming, 94% of all the DYS425 and
DYS439 markers and 73% and 74% of the DYS426 and DYS388 markers in Sykes full
data set, respectively, have a value of 12. This means that these markers do not have
sufficient spread and variability and are, in general, of limited use in
discriminating between haplotype patterns within this set of data.
Another interesting fact is the haplogroup distribution of
the data. Due to lack of haplogroup
designations in the original data, Athey’s haplogroup calculator was used as a
proxy to classify each haplotype. With
several haplotypes being removed because of missing data, Table 1 shows
a comparison of the results of the haplogroup calculator with Sykes published
“Clans.”
In this table, percentages shown in the middle column
reflect the output of Athey’s calculator while the percentages in the far right
column correspond to the breakdown published by Sykes in Appendix C of his
book.
It is clear from this comparison that Athey’s calculator
appears to classify the data in similar proportions as Sykes Clans and thus one
may infer the underlying meaning of Sykes’ clan nomenclature.
In addition, it is important to
note that Syke’s Clan categorizations were not based primarily on single
nucleotide polymorphism (SNP) testing.
As stated in the italicized quote above, “Y-chromosomes were assigned to clades, largely by RFLP analysis, these
assignments are not reported here as they do not necessarily correspond to the
SNP-based system recommended by the Chromosome Consortium.”
Finally, one should understand the regional borders that
Sykes uses in his study. Since the
purpose of this analysis is to draw geographic inferences, we are limited in
our insight by the definition of the geographic areas from which the data is
collected. Figure 1 shows the regional
borders that are coded in the OGAP data.
Table 1. Calculated
Haplogroups vs. Sykes Clans
Step 2 – Extracting R1b (Oisin) Data
The 1625 haplotypes identified as R1b in the previous step
were extracted from the data set. These
included all of the R1b haplotypes shown in Sykes Table 1, plus those for Ireland.
It is also interesting to note that when only R1b is considered, the
frequency of markers DYS426, DYS388, DYS439 and DYS425 with a value of 12 rises
to 97% in this subset of the database.
This further reinforces the notion that these markers (or the lack of
them) are not critical to the analysis of this set of data.
Figure
1. Regional Borders Used in the OGAP
Analysis to Classify Individuals
The first thing that was done to better understand the
data was to identify modal haplotypes for each region as a descriptive view of
the R1b data set. However, examination
of the modal haplotypes for the individual regions was not informative because
all regions and the full data set matched the standard Atlantic Modal
Haplotype.
A view of the R1b data in the form of a connected graph –
as shown in Figure 2 – shows a high degree of “cubism.” By this I mean a high degree of nodal
interconnectivity among the data points that results in opposite vertices
“washing out” differences in the data.
Clearly, data analysis based upon unique combinations of
markers (i.e. haplotypes) instead of individual markers would be necessary.
Figure
2. Network Analysis of the Top Twenty
R1b Haplotypes
Step 3 – Analysis of Haplotypes
Since descriptive statistics tended to “average out”
differences in the data, other methods were needed to identify patterns and
analyze the data. To do this the most
common haplotypes were identified and two methods of analysis were
performed. Appendix A shows the
haplotypes in the OGAP data. The OGAP
haplotypes roughly follow the frequency distribution of YSearch and in McEwan’s
(2007) groups if one takes into account that the OGAP data is light on Irish
samples in comparison with these other sources.
The OGAP designations in this table were assigned
sequentially in decreasing frequency of occurrence. The OGAP numbers from this table will be
referenced throughout the remainder of this report.
Two types of analysis were conducted to identify patterns –
affinity analysis and network analysis.
For affinity analysis of the haplotypes, an Excel
spreadsheet was developed to look for patterns and anomalies in the data. An algorithm was created that took as an
input the 10 marker values for a haplotype or signature to be reviewed and then
compared that haplotype to the R1b subset of the database. The algorithm calculated the genetic distance
and reported back the number of perfect (i.e., zero distance) matches by OGAP
region. To account for the differing
level of sampling in each region (e.g., small for Ireland and large for Northern England), the resulting number of matches
was normalized to frequencies in each region.
Table 2. Example of Identifying
OGAP8 Affinities
For example, haplotype OGAP8 which is generally considered
the quintessential Irish haplotype has 34 perfect matches in the database. Because Ireland was not extensively sampled, the
OGAP database has twice as many hits for Grampian as for Ireland. However, when we adjust the
Grampian and Ireland percentages for the nearly six
times greater sampling of Grampian, we see that the greatest geographic
affinity for the OGAP8 haplotype is Ireland.
This calculation is shown in Table 2 above. The OGAP8 results with decreasing presence in
Ireland, Argyll, and the Hebrides also corresponds with what is
generally understood to be the geographic distribution of this haplotype.
This analysis was repeated for the top 20 haplotypes in
the OGAP data. These haplotypes, which
cover 60% of all OGAP samples, appear sufficient to identify major regional
affinities. Analysis of additional haplotypes would be increasingly subject to
sampling error.
The results of the analysis of the top 20 haplotypes are
shown in Table 3. It should be
noted that in this table negative values were removed to reduce the clutter and
significant geographic anomalies were color coded to aid in identifying
tendencies. Finally, especially
interesting results were boxed to help in latter discussion in this paper.
License was also taken in the reordering of rows and
columns of the table in an attempt to group similar haplotypes and close
regions. While such analysis is called
“affinity analysis” and can be conducted mathe-matically, this analysis was
done manually to better allow for subjective considerations of the data.
The second type of analysis that was performed was network
analysis. This analysis which is common
in the genetic sciences was conducted using the Fluxus Networking program
version 4.2.0.0. Figure 3 shows
the results of the network analysis for the top twenty OGAP haplotypes.
It should be noted in Figure 3 that nodes have been relocated
and line length changed for increased readability. Nodes have also been colored to reflect the
regional affinities identified in Table 3.
Conclusions
Analysis of the Oxford Genetic Atlas Project data has
yielded interesting results. The combination
of the geographic affinity results shown in Table 3 and the network analysis
results shown in Figure 3 are synthesized in Figure 4. In this graphic, key haplotypes with strong
regional affinities were placed in their rough geographic perspective. No attempt was made to force every haplotype
somewhere on the map as it is obvious that at this level of analysis, some
haplotypes are pervasive and ubiquitous and not easily generalized to a single
geographic region.
Table 3 – Haplotype Affinity by
OGAP Region
Once located, haplotypes that differed by a single
mutation were connected with lines. Figure
4 reflects the general interconnectivity resulting from the network analysis
of Figure 3. The lines in Figure
4 should be thought of as one possible path of migrations – not necessarily
the only path. The interconnections
shown in Figure 4 are not based on any individual mutation rates. The interconnections shown in Figure 4
are based upon the occurrence of mutations, the principle of parsimony, and the
general south-to-north flow of R1b discussed by Sykes and Oppenheimer. Parsimony, in this case, reflects the
generally acknowledged flow from higher concentrations of haplotypes to lower,
more diffused concentrations.
Figure 4. Geographic Patterns of R1b in the British Isles
Some of the observations and conclusions of this analysis
are as follows.
1. The methodology clearly identified and quantified what has
been previous called the Irish subclade.
Whether called the Irish Modal Haplotype or the “Ui Neill haplotype” as
in the Trinity College paper,
the North-West Irish Haplotype by David Wilson and fellow researchers,
or R1bSTR19Irish as defined by McEwan (2007),
this corresponds to OGAP8 in the Oxford data. In addition, this analysis also identifies
the strong migration of OGAP8 to the Argyll area. This migration may be explained by the Dal
Riada migration
or as Sykes notes “Ar-gael was coined by
the kings of Dal Riada for their three colonies – Mull, Islay, and Kintyre peninsula.”
On page 217 he continues, “There
has certainly been a substantial settlement at some time from Ireland in the recent past and the Irish infiltration into the west of Scotland is almost certainly the signal of the relocation of the Dal Riada
from Ulster to Argyll in the first millennium.”
2. Similarly OGAP10 which shares the DYS390=25 allele
is clearly identified as an Irish haplotype.
This haplotype also shows up strongly in the Hebrides which supports Sykes and
Oppenheimer’s general conclusion of a Mesolithic northern migration along the
coast of Ireland.
3. Interestingly, OGAP5 is a very prevalent haplotype
that also shows up predominately in Ireland.
What is interesting is that this haplotype has DYS390=24 instead of the
generally accepted value of 25. This
haplotype, which has only one difference from the AMH modal (DYS389-1 = 14) may
be a transitional haplotype.
4. OGAP19 is interesting in that it shows an extreme
correlation with both Ireland and the Scottish Highlands.
Because of the low sampling of Irish participants, it is best to be
cautious about drawing strong conclusions concerning Irish population genetics
but clearly OGAP8, OGAP10, and OGAP19 show the strong ties that are shared by Ireland and the Highland area of Scotland.
It is the migration of these haplotypes westward during the first
millennium that make the Scottish to Irish migration in the 1700s so difficult
to isolate genetically. Specifically
Sykes writes, “There is a basic
similarity between Irish and Pictish Y Chromosomes which, incidentally, makes
it almost impossible to detect any genetic effect of the Ulster Plantations.”
5. OGAP4 is particularly intriguing. It is ubiquitous across all areas of Scotland and exceptionally strong in
Grampian, Tayside, and Strathclyde. If
we discount the Irish influx in Argyll and the Hebrides, it is also among the strongest
haplotypes present in these regions. It would not be too much of a stretch to
label OGAP4 the quintessential Scottish haplotype and the single closest
identifier to whatever is considered the indigenous Scottish population. Sykes and Oppenheimer both write that the
Picts were as close as anything to the indigenous population of Scotland. Sykes states on page 217,
6. OGAP6 is prominent in Argyll and the Hebrides.
It is also probable that OGAP6 is what Sykes was referring to when on
page 209 of his book he writes, “The
Argyll Y Chromosomes were much more like their Hebridean counterparts than those
in the Highlands … Argyll Y chromosomes are in between the Irish and the
Pictish values… Argyll has a low gene-sharing score with Grampian, even after
the Norse component has been subtracted.”
The first part of this statement can be observed by looking at the
affinities for OGAP6. The second part of
the statement further reinforces the notion that OGAP4 represents Pictishness
since the OGAP6 (Argyll) is a connection of OGAP4 to the Irish constellation in
Figure 3.
7. OGAP9 and OGAP11 both show an affinity for both the
Northern Isles and the Borders regions.
This affinity is distinctive, but the author is unqualified to venture a
theory that might explain this geographic discontinuity.
8. OGAP13 and OGAP17 both show a clear affinity for Northern England.
9. OGAP14, OGAP16, and OGAP20 all show a common
regional affinity. Though their presence
in Tayside is very slightly higher than in Central England, the author believes that they
are probably better categorized as English rather than Scottish. It is theorized that this signature probably
migrated northward along the east coast of England, but the author is at a loss as
to why there isn’t a greater presence of these haplotypes in the Borders, Northumbria, and Northern England.
Perhaps more historically astute researchers might provide a theory.
10. OGAP7 seems to be most prevalent
in Southern
England. OGAP12 has a stronger presence on the Borders
area, but similar to the comment above, it is speculated that this may have
been a northward migration but again the author can’t explain the apparent
geographical jump.
11. The core haplotypes for the full British Isles are OGAP1, OGAP2, and OGAP3. By their nature these haplotypes are
considered oldest and progenitors of the British R1b lines. This means that these haplotypes by their
very nature should be, and are, much diffused.
OGAP1 has a slight predisposition to Scotland and has its greatest
concentration in Wales. OGAP2 has a slight affinity for Ireland, and OGAP3 has a slight affinity
for Southern
England. It is unclear if much more can be said about
these haplotypes but with them one might see the original immigrants to the
Isles and their possible paths along the western and eastern coasts.
12. Wales is interesting. Sykes on page 239 makes the statement that Wales has the lowest diversity of R1b
variety than anywhere else in the Isles.
However this analysis shows that Wales has no predominant haplotype and
a relatively even distribution of other haplotypes used in this analysis.
13. Like Wales, the Isle of Man is at a similar crossroads. However, rather than the diversity of Wales, the Isle of Man shows a strong presence of three
haplotypes. Looking at the geographic
context of Figure 4, it can be seen that the Isle of Man was populated by OGAP10 eastward
from Ireland and westward by OGAP13 from
northern England and OGAP16 from central England.
Summary
Through the analysis of Sykes’ OGAP data, this study has
provided a means linking DNA results to haplotypes and conclusions in Sykes’
book, “Blood of the Isles.” The study
has confirmed Sykes’ interpretation of the data, and hopefully, provided a
means for other researchers to further validate and extend his work. The study both confirmed some subclades
identified by Sykes as well as identified some new subclades worthy of further
research. Key subclades that the study
posits and which are defined by Sykes include those of the Picts and the Dal
Riada Celts.
Picts – It is asserted that OGAP4 best
represents the Pictish ancestry of Scotland.
While there is no fundamental genetic difference between the Picts and
the Celts, i.e., both being R1b, they are both from the same mixture of Iberian
and European Mesolithic ancestry that forms the Pictish/Celtic substructure of
the Isles.
Dal Riada Celts – When considered in a narrow
genetic sense, the Gaels of Ireland, as identified by the DNA signature of
OGAP8, are as close as any group to being considered the root line and
forbearers of Celts of today.
When present in Scotland, it is suggested that OGAP8
represents the signature of the Dal Riada Celts. However, more generally, it should be clearly
noted that the term Celtic has taken on a near mystical meaning and now more
commonly refers to cultural, linguistic and regional patterns that have superseded
any genetic definition. Sykes says it
best when he states. “Overall, the
genetic structure of the Isles is stubbornly Celtic, if we by that mean descent
from people who were here before the Romans and who spoke the Celtic
language. We are an ancient people
…genetically rooted in the Celtic past.
The Irish, the Welsh, and the Scots know this, but the English sometimes
think otherwise.”
Several interesting clusters were identified that show
geographic affinities but discontinuities.
Scotish clusters OGAP9 and OGAP11 have a strong presence in the Borders
as well as the Northern Isles. English
clusters OGAP14, OGAP16, and OGAP20 show a predisposition to both Central England as well as Tayside. These and other geographic affinities warrant
further research.
Electronic Database
Information
Capelli, C. et al. 2003 Data Set: http://freepages.genealogy.rootsweb.com/~gallgaedhil/Capelli.htm
Oxford Genetic Atlas Project (OGAP) Data
Set:
http://www.bloodoftheisles.net/results.html
John
McEwan’s R1b Haplotypes:
http://www.geocities.com/mcewanjc/p3modal.htm
Whit
Athey’s Haplogroup Predictor:
http://home.comcast.net/~hapest5/index.html
Athey TW (2006) Haplogroup prediction from Y-STR values using
a Bayesian allele-frequency approach. J
Genetic Genealogy, 2:34-39.
McEwan
J (2007) Phase 3 analysis: Ysearch 37
STR modal summary and analysis tables (web site).
Oppenheimer S (2006)
The Origins of the British - A Genetic Detective Story, Constable
and Robinson, London (ISBN 1-84529-158-1). Published in the U.S. by Carroll and Graf, New York.
Sykes B (2006) Blood
of the Isles: Exploring The Genetic Roots of Our Tribal History. Bantam Books.
Published in the U.S. as Saxons, Vikings, and Celts:
The Genetic Roots of Britain and Ireland, by W. W. Norton, New York.
Appendix A – OGAP
Haplotypes
The 1625 data points that comprise the R1b data set
include 291 separate haplotypes.
However, 50% of the data can be accounted for with only 10 haplotypes;
60% by 20 haplotypes; and 68% by 30 haplotypes.
In addition, haplotypes beyond the top 30 have only single-digit frequencies
compared to the most frequent – the Atlantic Modal Haplotype (AMH), which
occurs 262 times in the data.
Below are the top 50 haplotypes in order of descending
frequency. These are number OGAP1
through OGAP55 for reference in this study and represent all haplotypes that
occur more than 5 times. The
distribution of these haplotypes in Sykes’ OGAP data and in YSearch (www.ysearch.org) as of December 2006) is
shown on the right hand side of the chart.
Also, a mapping of John McEwan’s (2007) R1b subclades is
shown on the left hand side of the chart.
The letters designating the McEwan group refer to the groups described
in Appendix B that were assigned when McEwan individual haplotypes were grouped
together to reflect the much smaller number of markers in the OGAP data.
Appendix B – McEwan’s R1b Haplotypes Reduced
By mining YSearch and collecting similar 37-marker
haplotypes into clusters, McEwan (2007) has identified a large number of R1b “types”
that comprise the world-wide scope of this data. It is interesting to consider how the Sykes
OGAP data relates to McEwan’s haplotypes.
However to compare the data, McEwan’s modal haplotypes had to be collapsed
into smaller groups to reflect the smaller number of markers in the OGAP haplotypes
(Refer to Appendix A).
The following is the reduction of the McEwan haplotypes to
10-marker haplotypes. In each case, letter
designations have been included here for the purpose of mapping the full set of
McEwan haplotypes (designated R1bSTR##) to the reduced set used in this
analysis (Letter Groups). While this exercise necessitates the loss of
considerable resolution in the McEwan haplotypes, the exercise is included here
to provide traceability to the analysis included in Appendix A.
In the case of academic researchers,
many keep tight control of their data.
It is hoped that in the future that scientific journal editors will
require the submission of supporting data, which they might even hold for a
period of time even after publication of analytic articles. Such submissions would ensure that important
research is fully documented even if it is years later when the article is no
longer at the forefront of everyone’s mind.
Address for
correspondence: Campbell@alum.mit.edu
