PreprintPDF Available

Response to Rzepecki et al., "Missing link in Late Antiquity? A critical examination of Hollstein’s Central European Oak Chronology"

Authors:
Preprints and early-stage research may not have been peer reviewed yet.

Abstract and Figures

In a new article in Dendrochronologia, Andreas Rzepecki with co-authors lift the lid on Ernst Hollstein's weak bridge over the Roman gap in the Central European oak chronology. This issue has been taboo since the bridge was accepted by academia. However, the authors do not deliver any scientific proof for their assertion that the bridge is still valid. Our analysis shows that the generally used confidence levels for dendrochronological matches are still far too low to point out an unambiguous synchronous position. And in cases when a strong confidence level can not be reached with dendrochronology, the use of less resolved methods like radiocarbon, or even worse historical considerations, is still regarded an adequate procedure.
Content may be subject to copyright.
Response to Rzepecki et al., draft, 2019-02-04, Page 1 of 16
Response to Rzepecki et al., "Missing link in Late Antiquity? A critical
examination of Hollstein’s Central European Oak Chronology"
Petra Ossowski Larsson* and Lars-Åke Larsson, Sweden
* Corresponding author: petra@cybis.se
Abstract
In a new article in Dendrochronologia, Andreas Rzepecki with co-authors lift the lid
on Ernst Hollstein's weak bridge over the Roman gap in the Central European oak
chronology. This issue has been taboo since the bridge was accepted by academia.
However, the authors do not deliver any scientific proof for their assertion that the
bridge is still valid.
Our analysis shows that the generally used confidence levels for dendrochronological
matches are still far too low to point out an unambiguous synchronous position. And
in cases when a strong confidence level can not be reached with dendrochronology,
the use of less resolved methods like radiocarbon, or even worse historical
considerations, is still regarded an adequate procedure.
Introduction
A new article by Andreas Rzepecki et al. (ref.1), including as co-authors the heads of
the German public dendrolabs in Trier, Cologne and Freiburg, attempts to re-
evaluate the dendrochronological link between the Roman complex and the Early
Medieval complex defined by the most important Central European oak chronology
published in 1980 by Ernst Hollstein (ref.2). This is an excellent initiative, because
our own investigations made ten years ago have shown that Hollstein's link is very
weak and hence that the Roman complex is floating (ref.3). A floating Roman
complex has possible implications not only for history, but also for e.g. climate
science which tries to synchronize climate events visible in tree-ring archives with
reported societal effects. Corrupt dendrochronology would lead to wrong conclusions.
Regrettably, the authors do not apprehend the meaning of "floating". Floating in
dendrochronology means that a partial chronology cannot be assigned an absolute
end date because it can not be linked significantly with an already absolutely dated
other chronology part. As long as this significant link has not been established with
dendrochronology, the floating chronology part could theoretically assume any
possible end date. It is of course possible to roughly confine the floating part on the
absolute time line by using radiocarbon measurement. But using historical
considerations to place the floating chronology in a narrow time frame, followed by
accepting the first and best weak dendrochronological match as the true one, is not
an appropriate practice.
This was pointed out already in 1981 by Mike Baillie when he wrote a review about
Hollstein's book (ref.4):
Response to Rzepecki et al., draft, 2019-02-04, Page 2 of 16
This raises the danger of circular argument, viz the sample is placed roughly in
time on historical/archaeological information, a match is searched for, and if an
acceptable match is found the resultant date will reinforce the original
information; such a procedure departs from the essentially independent nature
of dendrochronological dating.
Which did not prevent Baillie from going into the same trap when it came to the weak
links of the Belfast chronology.
The authors admit that Hollstein's original dendrochronological link is insufficient.
Furthermore they identify one of the mean value site curves making up the link as
misdated by 138 years. This means that Hollstein must have been almost clairvoyant
if he got it right anyway. Nevertheless the authors are firmly convinced that the link is
still valid and that the end date of Hollsteins' Roman complex does not need to be
changed. They explain this conviction with the statement, that "criticism of the
general accepted medieval timeline has already been disproved by various scientific
disciplines". Everything else would of course make bizarre nonsense of the written
history.
That the link is still valid can be proved only by demonstrating a sufficient
dendrochronological crossmatch. Because the only scientific dating method with
annual resolution known today which can be practized completely independent of
historical prerequisites is dendrochronology. What the authors mean by various
scientific disciplines is enigmatic. Most probably they refer to astronomy, which has
been used to place ancient observations of celestial phenomena handed down via
various media on an absolute time line. But as retrospective astronomy depends on
the availability of written historical reports which may be correct or retrocalculated or
even fake, you can not prove the correctness of the general accepted medieval
timeline with it. Read our extensive investigation about historical astronomy in ref.5.
The following analysis will investigate if Rzepecki et al. are able to demonstrate a
sufficient alternative dendrochronological link compatible with Hollstein's, or not.
Methods and data
We use CDendro (ref.6) as the tool for dendrochronological analysis and crossdating.
A number of well-tried algorithms for crossdating and normalization are available in
the program. We usually look at several methods simultaneously and require a high
score from all of them to accept a match. In this case we use the Hollstein
normalization as the sorting method and a requirement on overlap of 50 years when
crossdating, if not otherwise specified. Crossdating is done under interactive
supervision with graphical tools and quality check functions. About our acceptance
levels for crossmatches see ref.7.
We use the following mean value site curves (raw data is not available) in our
analysis:
Response to Rzepecki et al., draft, 2019-02-04, Page 3 of 16
Rzepecki
et al.
(ref.1,
Appendix A)
Dates
(ref.1)
Corr. coeff. curves
ref.1 compared to
ref.2
Hollstein,
digitized site
curves (ref.2)
Dates
(ref.2)
T_SPEI001 12 - 316
TATE 70 - 286
MB253 76 - 253
T_KOEL015 113 - 314
0.98 KDGB 232 - 314
T_KOEL007 149 - 336
0.95 KORB 149 - 336
T_OTTE001 158 - 315
T_GUST001 242 - 373
TRMB32 183 - 315
OB2B 401 - 550
F_BEFO001 250 - 504
T_MAIN023 260 - 394
F_ENNE001 266 - 510
T_BROI001503
267 - 503
0.96 BROI 129 - 365
T_KOEL016 274 - 511
0.97 KKNA 273 - 511
T_BEER001 284 - 568
0.95 BERL 284 - 550
F_FLAV001 306 - 567
T_ERPF001 410 - 643
0.89 ERP2 410 - 643
T_MAIN012 439 - 716
0.93 MB53 + MB16 439 - 716
T_TRIE003 442 - 670
0.86 *) TRAM 442 - 668
HUFI 450 - 606
T_STEI001 474 - 814
0.97 STEI 474 - 814
MB73 481 - 673
T_ALTP001 494 - 691
0.95 ALT1 494 - 691
T_MUEE002 521 - 683
0.99 MUES 521 - 682
Table 1: Mean value site curves used in this article. Many of the curves marked "T_" retrieved from
Appendix A of Rzepecki et al. (ref.1) are identical with Hollstein's drawn curves, digitized by us (ref.2).
Use the table for conversion. How well the curves were drawn and how well the digitisation worked is
demonstrated by the correlation coefficients which we got when comparing the curves of ref.1 (from
original datasets) and ref.2 (from digitzed drawn curves) with each other. The corr. coeff. is at least
0.89 and often much higher, which is generally better than the corr. coeff. normally achieved between
different measured radii of the same sample. The usefulness of our digitized curves has been
questioned by one of Rzepecki's co-authors (ref.12).
*) Hollstein apparently missed two rings in TRAM near to year 610
(http://www.cybis.se/forfun/dendro/hollstein/arenakeller2/index.htm, 2010), which has been corrected
in T_TRIE003. With this correction, TRAM fits against T_TRIE003 with corr. coeff. 0.95.
Appendix A of ref.1 is an Excel-file containing the ring-width data of 62 mean value
site curves as a matrix. It is easy to convert this file into a collection in Heidelberg
format (.fh) using CDendro, see Appendix A in this article. There are two small errors
in the matrix: the mean value curve T_KOEL00X has to be renamed to T_KOEL015
in order to be compatible with the article text. And the mean value curve T_BROI001
has to be redated to 503 in order to mirror the authors' opinion about the end date
which is actually used in the "new Late Antiquity bridging" (ref.1, Table 2).
Response to Rzepecki et al., draft, 2019-02-04, Page 4 of 16
Analysis of the alternative linkage suggested by Rzepecki et al.
Of the 62 mean value curves contained in Appendix A of ref.1, 12 curves are directly
involved in the link between the Roman complex and the Early Medieval complex.
The Roman complex is represented by four curves, T_SPEI001, T_KOEL015,
T_KOEL007 and T_OTTE001, which have a moderate fit towards each other (see
table 5). That these curves indeed belong to the Roman complex can be safely
demonstrated only if more material e.g. of the Hollstein chronology is added (see
table 7). The end date of the youngest curve, T_KOEL007, is 336 conventionally.
The other eight curves belonging to the Early Medieval complex have all start dates
before 336: T_GUST001, F_BEFO001, T_MAIN023, F_ENNE001, T_BROI001503,
T_KOEL016, T_BEER001 and F_FLAV001 (see figure 5).
Hollstein's original link between the Roman complex and the Early Medieval complex
of the European oak chronology, which still is regarded as valid, is described in
Figure 1 and Table 2.
Figure 1: Time line diagram for the mean value curves representing the original Hollstein link in the
new Late Antiquity bridging in ref.1.
Table 2: Crossdating quality at the conventionally assumed position. Displays corr. coeff., t-value and
overlap for the dendrochronological fit of each curve towards all others. Hollstein normalization used.
(An '*' indicates a very low correlation value though the overlap is not too low (corrCoeff < 0.2 and overlap > 40).)
T_KOEL007 is the youngest of Hollstein's site curves which can be safely
synchronized with the Roman complex. T_KOEL016 and T_BEER001 belong to the
Early Medieval complex. Their 3rd century start dates (see table 1) have been
confirmed with moderate probability in ref.1. We have never questioned these dates,
but they are impossible to confirm within the original Hollstein chronology.
The link between the two complexes, shown in the column for T_KOEL007 in table 2,
is far from significant - the t-values are low and the overlaps are short - and we could
already here stop reasoning.
But let's have a look at some additional material. There is only one more site curve in
the Hollstein chronology, T_BROI001, which can be related to the link directly.
Hollstein placed T_BROI001 at end year 365:
Response to Rzepecki et al., draft, 2019-02-04, Page 5 of 16
Figure 2: Time line diagram for the mean value curves representing the original Hollstein link
including T_BROI001365 at the position suggested by Ernst Hollstein.
Table 3: Crossdating quality at the conventionally assumed position, with T_BROI001365 included at
the position suggested by Ernst Hollstein.
In this position, T_BROI001 shows some kind of match only with T_BEER001, all
other comparisons do not result in the identification of any acceptable synchronous
positions. This led Rzepecki et al. to propose the alternative end year 503 for
T_BROI001:
Figure 3: Time line diagram for the mean value curves representing the original Hollstein link
including T_BROI001503 at the position suggested by Rzepecki et al.
Table 4: Crossdating quality at the conventionally assumed position, with T_BROI001503 included at
the position suggested by Rzepecki et al.
Maybe this looks a little bit better, but now we have to ask how this link is compatible
with the other Roman site curves which end around 315:
Response to Rzepecki et al., draft, 2019-02-04, Page 8 of 16
Again we look at quadrant 3, in table 6. By adding T_GUST001 and T_MAIN023 to
the Early Medieval complex the match towards the Roman complex has been
boosted, but only for T_KOEL007 which has two matches with t>4.5. All other
matches are under the acceptance criteria (t>4.5) stated in ref.1, and most are still
disastrous.
If we calculate a mean value curve each for the Roman complex and the Early
Medieval complex shown in table 6, the comparison of these two curves results in:
corr. coeff. 0.42, t=4.5 (Hollstein) and corr. coeff. 0.45, t=4.8 (Baillie-Pilcher), at 94
years overlap. Gleichläufigkeit (GLK) 0.69. But if we exclude T_KOEL007 from the
Roman complex, the match between the blocks collapses: corr. coeff. 0.25, t=2.2
(Hollstein) and corr. coeff. 0.25, t=2.1 (Baillie-Pilcher), at 74 years overlap. GLK 0.65.
I.e. the solution depends heavily on T_KOEL007 alone.
An alternative linkage within the Hollstein chronology
We'll now show you an alternative link between the Roman complex and the Early
Medieval complex within the Hollstein European oak chronology, also including
T_KOEL007 but using some more of Hollstein's original site curves, see figure 6 and
table 7. This link implies that the Roman complex has to be dated 207 years nearer
to our time, i.e. T_KOEL007 assumes the end date 543 instead of 336.
Looking at quadrant 3 of table 7, there are now five acceptable matches especially
for T_KOEL007 (with the acceptance criteria (t>4.5) stated in ref.1).
If we similar as above calculate a mean value curve each for the Roman complex
and the Early Medieval complex in table 7, the comparison of these two curves
results in: corr. coeff. 0.48, t=6.4 (Hollstein) and corr. coeff. 0.40, t=5.2 (Baillie-
Pilcher), at 142 years overlap. Gleichläufigkeit (GLK) 0.71. If we exclude T_KOEL007
from the Roman complex, the match between the blocks does not collapse
completely: corr. coeff. 0.43, t=5.2 (Hollstein) and corr. coeff. 0.34, t=3.9 (Baillie-
Pilcher), at 122 years overlap. GLK 0.70.
This alternative link, right away present in the original Hollstein data, is actually much
better than the link Rzepecki et al. (ref.1) assert being a proof for the correctness of
the general accepted medieval timeline!
Response to Rzepecki et al., draft, 2019-02-04, Page 10 of 16
Discussion and Conclusions
The new article by Andreas Rzepecki et al. (ref.1) succeeds to extend the Early
Medieval complex of Ernst Hollstein's oak chronology to the middle of the 3rd century
CE. However, the authors are not even near to demonstrating a significant and
credible link to the Roman complex which contains site curves archaeologically
anchored in Roman history. The overlaps making up the link are short and most of
the correlation coefficients and t-values are really low or even negative which hints at
errors in the overlapping tails of the curves. Moreover, the link depends on one single
curve, T_KOEL007. If this curve is omitted the link collapses completely.
The correlation analysis of the mean value curve of the Roman complex compared
with the mean value curve of the Early Medieval complex reveals that the acceptance
criteria stated by the authors (overlapping above 80 years, Gleichläufigkeit (GLK)
>60% and t-values (Hollstein and Baillie-Pilcher) >4.5) are met with the smallest
possible margin. Even though the authors are content with a t-value of 4.5, as they
state that this value is chosen stricter than usual, this value is far from pointing out
only significant synchronous positions. We will demonstrate this with T_BROI001 as
an example.
We calculate a mean value curve from all site curves in ref.1, Appendix A without
T_BROI001. The resulting mean value curve is 839 years long, dated 12 to 850, and
becomes our reference curve. Then we match T_BROI001 against it.
Figure 7: Correlation coefficients (upper, red part of diagram) and t-values (lower, blue part of diagram) for all
possible positions of T_BROI001 against a 839 years long Late Antiquity reference. Hollstein normalization,
requirement on overlap 50 years. Threshold t=4.5 marked by grey line. Use the magnifying function of your
reader or browser to see details.
There are two possible synchronous positions with a t-value above 4.5, 503 and 445,
and one position with t=4.3, 365. But only one of these can be the true one. Hollstein
chose 365 and Rzepecki chose 503, but why not 445? You can not really tell which
position is the true one, not even with visual inspection, because all three are about
equal and only slightly above the noise level which apparently is at about t=3.5. For
some reason the true synchronous position is not discriminated.
If you in this situation guess on the wrong position, you will get a misdated site curve.
Maybe that does not matter too much if you are going to date an archaeological
sample. But if you practise this in master chronology building it could compromise the
entire work. This is why we have dismissed T_BROI001 when we analyzed the
Roman link in the Hollstein oak chronology (ref.3).
Response to Rzepecki et al., draft, 2019-02-04, Page 11 of 16
Now, what is a good t-value? In Belfast t=3.5 was adopted as a "threshold for
significant matches" when dealing with oak series longer than 100 tree-rings and
using the Baillie-Pilcher normalization method before the calculation of the correlation
coefficient (ref.8). This threshold was assumed to correspond to a probability of 1
false match per 1000 mis-matches, i.e. about three standard deviation levels in a
normal distribution.
A probability of 1 false match per 1000 mis-matches means that each case like the
one shown in figure 7 with about 1000 possible overlapping positions is expected to
generate one false significant match above t=3.5. This does not matter if there also is
a strong true synchronous position present, but could lead to erronously crossdated
site curves if the sample has nothing to do with the reference curve either in time or
place, or if there is an error in it.
Therefore we oppose calling t=3.5 the "threshold for significant matches". In signal
processing we would call it the "noise level". We strongly propose that there should
be a second, higher threshold for the acceptance of a match and that this threshold
should have a clearance of at least two t-units above the noise level. We are fully
aware that it might be unrealistic to request such high acceptance levels when
dealing with the dating of archaeological wood, but when building master
chronologies we should search a firm level of confidence. Using pre-dating with less
resolved methods like radiocarbon, or even worse historical considerations, in cases
when such confidence level can not be reached dendrochronologically, is what we
call a "lowering of the guards". Regrettably, it seems that this was exactly what
happened when the final gaps in the European oak chronologies were closed (ref.3).
Having said this, we could refine our question: what is a good t-value if we want to
rely on dendrochronology alone? We will get an approximate empirical answer if we
take a long (4000 years) reference curve and try to match a large sample collection
(more than 500 samples, >100 years long) against it. The time range of the sample
collection must not coincide with the time range of the reference (because we want
all matches to be false), but we choose to use oak samples from a narrow
geographical area. The t-values for the best matches from this test will indicate the
level at which we still get false matches. Thus for correct crossdating which relies on
dendrochronology alone, we need t-values well above these levels.
The results from the test described above differ slightly for various normalization
methods.
Irish
Oak
t-values
Normalization method best 2nd 3rd
P2YrsL 5.4 5.3 5.3
P2Yrs 5.4 5.3 5.3
Hollstein 5.5 5.4 5.3
Baillie-Pilcher 5.6 5.5 5.4
Cross84 6.0 5.7 5.6
Besançon Index E 5.9 5.8 5.8
Mean sliding frame 5.3 5.3 5.2
Table 8: Best three matches (t-values) of a 500 samples collection (AD-time, sample length >100
years) against a 4000 years reference chronology (BC -5000 to -1000) for Irish oak using various
normalization methods available in CDendro. All matches are false. Minimum required overlap 70
years, minimum required corr. coeff. 0.35.
Response to Rzepecki et al., draft, 2019-02-04, Page 12 of 16
The somewhat sobering figures in table 8 hint at a threshold for the acceptance of a
match of at least t=6 for oak, far above the noise level of t=3.5 and even the stricter
acceptance level of t=4.5 as adopted by Rzepecki et al.
Therefore the alternative match which would set the Roman complex 207 years
nearer to our time is deeply disturbing. Here we reach t=6.4 at an overlap of 142
years. We demonstrated this link at a round table discussion in Bibracte in 2010
(ref.9). The conclusion of this seminar was that the Celtic and Roman block of the
Hollstein reference (1980) is floating, and that the original Hollstein link ought to be
re-investigated. Regrettably, this came to nothing and when we tried to publish our
results together with even more worrying new-found details our manuscript was
rejected in three peer review rounds (ref.10). This is why we finally self-published on
ResearchGate and Academia.edu.
We actually do not think that the 207-years-match is the full truth either. It is still too
weak to be convincing, and also this link depends to a certain extent on T_KOEL007
which we suspect has an error in its younger end. We have tried unsuccessfully to
get more raw material of the Roman Rhine bridge in Cologne, because we know that
about 90 timbers of that structure were retrieved from the riverbed (ref.11). Hollstein's
T_KOEL007 is a mean value curve of only four timbers from that site. Therefore it is
disappointing that Rzepecki et al. do not question the original measurements made
by Ernst Hollstein but use them unencumbered in their new link, see table 1. Though
it is interesting to see that samples bridging the gap still seem to be rare.
However, there are much more indications that the original Hollstein link is in trouble
than just weak dendrochronology in the actual Hollstein master. Rzepecki et al.
certainly cite our ref.3 but ignore that we in that article also claim that:
all other published European oak chronologies too fail to bridge the Roman
gap
there is an error in the Belfast chronology which challenges its internal linkage
there is an even more deeply disturbing significant teleconnection against the
Scandinavian pine masters suggesting that the European Roman oak complex
conventionally is dated too old by more than 200 years
Moreover, already Ernst Hollstein was unhappy with his own dendrochronological
linkage which assigned the building year 336 to the Rhine bridge in Cologne. This
bridge is mentioned as already founded in a panegyric on Constantine the Great
dated to year 310, 26 years before the foundation timbers were cut according to the
dendrochronological analysis (ref.2, page 74). Hollstein also complained about the
bridge over the Moselle in Trier which got a safe dendrochronological felling year 71
within the Roman complex for one of the large foundation pillars. However, that
bridge played a role already during the Batavian uprise year 70 as narrated by
Tacitus and the bridge was therefore certainly completed before that year (ref.2,
page 11).
English Heritage even felt bound to declare Tacitus a liar as his narrative about
Agricola, his father in law, did not comply with the dendrochronological dates of
timbers from the different construction phases of the Roman fort at Carlisle (ref.13).
All this indicates that the dendrochronological and historical Roman complexes might
be mis-aligned, the dendrochronological complex giving a few years too young dates
Response to Rzepecki et al., draft, 2019-02-04, Page 13 of 16
compared to the historical complex. Read our hypothesis in ref.5, especially
Appendix E, about how this could be explained.
Andreas Rzepecki and his co-authors definitely deserve credit for lifting the lid on
Ernst Hollstein's weak bridge over the Roman gap. This issue has been taboo since
the bridge was accepted by academia. However, after a brief peek they put the lid in
place again saying that they did not find anything spectacular and everything is fine.
This is a bit embarrassing as we have discussed above, because they deliver no
proof for such an assertion.
More embarrassing is their comparison of the final "Late Antiquity bridging" with an
unpublished master chronology from southern Germany, said to be completely
independent of the Hollstein chronology. But there is no such thing, all European oak
chronologies are perfectly synchronized with each other, the Hollstein chronology
being the template for the rest because it was first. Some secondary chronologies
have gaps while others are boasting with multiple replication for every year. But this
is not difficult to achieve with the low acceptance criteria which apparently are
generally applied. However, there is no European oak chronology which has a
documented free bearing bridge over the Roman gap.
Instead of invoking the chronology from southern Germany, the authors could just as
well have consulted the old Becker chronology (ref.14). That one gives corr. coeff.
0.48, t=15.8 (compared to corr. coeff. 0.58, t=16 as stated by the authors for the
unpublished master). As long as a significant dendro bridge between Early Medieval
complex and Roman complex has not been demonstrated for the chronology from
southern Germany (which is quite unlikely as the chronology is unpublished), we
have good reasons to regard this comparison as a classical case of circular
argument.
And even more embarrassing is the fact that a paper with such shortcomings passed
through the peer review in a top scientific journal.
After this verbiage and all this "heresy", which scientific journal would publish our
article? This is why we also this time choose to "self-publish" our manuscript on
ResearchGate. The advantage is that you can read it instantly, but you have to do
the peer review yourself.
Response to Rzepecki et al., draft, 2019-02-04, Page 14 of 16
References
1. Rzepecki A, Neyses-Eiden M, Frank T, Diethelm B, Herzig F, Tegel W (2019). Missing link in Late
Antiquity? A critical examination of Hollstein’s Central European Oak Chronology, Dendrochronologia
vol. 54, pages 20-28. https://doi.org/10.1016/j.dendro.2019.01.003
2. Hollstein, E. Mitteleuropäische Eichenchronologie. Verlag Philipp von Zabern, Mainz am Rhein
1980. Digitized mean value site curves download see Appendix B.
3. Ossowski Larsson P. & Larsson L.Å. (2015). Dendrochronological dating of Roman time.
ResearchGate DOI: 10.13140/RG.2.1.5129.6806.
https://www.researchgate.net/publication/275083761_Dendrochronological_Dating_of_Roman_Time
4. Baillie, M. (1981). Mitteleuropäische Eichenchronologie. By Ernst Hollstein. Trierer Grabungen und
Forschungen, Band XI, 273pages, 67 illustrations. Phillip Von Zabern, Mainz am Rhein, 1980.
Proceedings of the Prehistoric Society, 47, 356-356. doi:10.1017/S0079497X00009245
https://www.cambridge.org/core/journals/proceedings-of-the-prehistoric-
society/article/mitteleuropaische-eichenchronologie-byernst-hollstein-trierer-grabungen-und-
forschungen-band-xi-273pages-67-illustrations-phillip-von-zabern-mainz-am-rhein-
1980/ABBA5907EC4DD30E0CA0805A36C9ECC9
5. Ossowski Larsson P. & Larsson L.Å. (2016). Astronomical dating of Roman time. ResearchGate
DOI: 10.13140/RG.2.1.2703.3363.
https://www.researchgate.net/publication/296060902_Astronomical_dating_of_Roman_time
6. CDendro available at: http://www.cybis.se/forfun/dendro
7. Ossowski Larsson P. & Larsson L.Å. (2018). Validation of the supra-long pine tree-ring chronologies
from northern Scandinavia. ResearchGate DOI: 10.13140/RG.2.2.18643.89124.
https://www.researchgate.net/publication/322662787_Validation_of_the_supra-long_pine_tree-
ring_chronologies_from_northern_Scandinavia
8. Baillie, M. G. L. (1982). Tree-Ring Dating and Archaeology. (Croom-Helm, London)
9. http://www.disci.unibo.it/it/risorse/files/0-Programme_du_seminaire.pdf, or
http://www.webcitation.org/75uyPWfVg to download a cache copy of this page (cached 2019-02-03).
10. http://www.cybis.se/dendro/dendro-audit/peer-review
11. http://dendro.phil-fak.uni-
koeln.de/fileadmin/ufg/pdf/labore/dendroarchaeologie/projekte/roemerzeit.pdf, or
http://www.webcitation.org/75TDrd2Mt to access a cache copy of this page (cached 2019-01-16).
12. ITRDB Forum 26 Aug. 2008: Dear all, It seems for me quite pointless to discuss data and possible
errors within Hollstein's seminal work from 1980. The mean curves of the sites published by Hollstein
are hand-drawn and will serve in the first line for visualisation. Simple digitisation of these curves in
order to gain absolute values is highly problematic. It will be certainly not appropriate to use these data
for checking, tests and as premises (invented 300 years). On the contrary, raw data as measured
values will be required.
13. Dendrochronology. English Heritage, June 2004. https://historicengland.org.uk/images-
books/publications/dendrochronology-guidelines
14. Becker, B. 1981. Fällungsdaten Römischer Bauhölzer anhand einer 2350-jährigen Süddeutschen
Eichen-Jahrringchronologie. Fundberichte aus Baden Württemberg No.6, 369-386.
Response to Rzepecki et al., draft, 2019-02-04, Page 15 of 16
Appendix A: How to get Rzepecki et al.'s matrix data into CDendro
Click on the link “Appendix A. Supplementary data” in the online version of ref.1.
Save this file in Text format (tab delimited) as e.g. “Rzepecki.txt”.
Open the txt-file with a text editor (e.g. Notepad) and remove the first three
“unexpected lines” from the text window. They contain lots of ending Tab-characters,
be careful to remove them all so that the very first line starts with “Year
F_BEFO001 F_BE…”. Then save and close the file.
In CDendro select the command “Collections/Import and convert tabular
data…/Import a multi column table of data” which opens a “command window” where
you can specify the data format.
1. Click “Paste from a text file” and select your .txt-file.
2. See that “Field delimiter character” is set to Tab character.
3. See that “Find sample identities in top row” is checked.
4. See that “Find End-Year in leftmost column” is checked.
5. Click the button “Analyze data from textbox now!”. If you later make any changes
to the textbox, do remember to then again click this “Analyze” button. Max row index
will show “839” and Max column index “62”.
If you now set both RowIndex and ColumnIndex to 0 (zero) and click “Get cell data”,
the text “Year” should pop up on the line “Cell data found”.
Response to Rzepecki et al., draft, 2019-02-04, Page 16 of 16
6. See that the “Scale factor” is set to “1”, meaning that the values from the matrix will
go unchanged into your new .fh- file.
7. Click “Try to convert data read into a new .fh collection file” and then select a
temporary filename.
Close the “command window” and open the new .fh collection file with
“Collections/Open menu for collections…/Open Heidelberg collection (.fh)” and find
your new temporary collection file.
In this special case all collection members have lengthy sequences of zeroes before
and after their real data. You have to remove them by saving the new collection file
once again with Collections/Save current collection/…as Heidelberg (.fh) formatted
file…” with e.g. the name “Rzepecki.fh”. (You can then remove the temporary file
which is not needed anymore.)
Appendix B: Digitized mean value site curves from Ernst Hollstein,
Mitteleuropäische Eichenchronologie (ref.2)
Ernst Hollstein's relative ring width data retrieved from drawn curves are available for
research purposes. Download here:
https://www.researchgate.net/publication/330761703_Digitized_mean_value_site_cur
ves_from_Ernst_Hollstein_Mitteleuropaische_Eichenchronologie
or with a lot of comments here:
http://www.cybis.se/forfun/dendro/hollstein/index.php?hollstein1
... A recent attempt to declare Hollstein's bridge over the Roman gap as still valid gave us the opportunity to demonstrate how weak this bridge actually is (ref. 4). ...
Preprint
Full-text available
This article is about the historical consequences of our scientifically reinforced hypothesis that the West-Roman empire is conventionally dated some 232 years too old. We offer an alternative interpretation of some Roman heirlooms retrieved from the grave of the Frankish king Childeric, and from a Japanese grave dated to the late 5th century.
Preprint
Full-text available
This article is about a rarity: radiocarbon dates of timbers archaeologically anchored in West-Roman time which are also dated by dendrochronology. The surprising but apparent trend is that the radiocarbon dates are a large number of years younger than the dendro dates. This strongly supports our hypothesis that West-Roman history and archaeology are conventionally dated too old by more than two hundred years, and that European dendrochronology was adapted to this error already in its early period.
Research
Full-text available
We empirically demonstrate that acceptance thresholds for a dendrochronological cross-match of at least t=6 for oak and at least t=7 for pine are required when constructing independent tree-ring master chronologies. These thresholds are far above the "thresholds for significant matches" of t=3.5 or t=4 adopted forty years ago, which are still regarded appropriate for the dating of archaeological samples. We also validate the supra-long Scots pine chronologies from northern Scandinavia using these tightened criteria and our crossdating software CDendro. The Finnish and Swedish pine chronologies were built by two concurrent teams of scientists working with wood from different places and with different methods. Both teams arrived at the same conclusion regarding the overall dendro signal for the past 7500 years. Our reassessment confirms this conclusion, and demonstrates that both teams worked on a firm level of confidence when accepting dendrochronological matches. This validation is vital for the credibility of our hypothesis about general errors in the European oak masters, errors which were probably caused by using pre-dating with other methods in cases when sufficient dendrochronological confidence levels could not be reached.
Research
Full-text available
Published or otherwise available European oak tree-ring chronologies archaeologically anchored in Roman time are all separated from early medieval chronologies by a severe timber depletion in late antiquity. Our recent dendrochronological study shows that this gap probably is unnecessarily wide because the Roman dendro complex as a whole appears dated too old by 218 years. The subject of the here presented astronomical study was to investigate if there is additional scientific support for such a mistake which would mean a large calendar error in the Christian era. Our results indicate that the Christian era was inflated with 232 years already when it was invented. This was done by backdating West-Roman and related history by means of astronomical retrocalculation after the western part of the Roman empire had declined. A remarkable result of our astronomical study is that the postulated astronomical/ historical error (232 years) appears to be offset by 14 years from the dendrochronological error (218 years). This means that, if we are right, then all current dendrochronological dates within the Roman time complex are given 14 years too young. According to our interpretation, the 14 years offset was caused by an improper synchronization of the Roman dendro complex towards Roman history done more than 30 years ago.
Research
Full-text available
Based on published and otherwise available tree-ring data, we have analyzed the dendrochronological support for the current dating of Roman activities in western Europe. Manuscript rejected by Tree-Ring Research, details of peer review see: www.cybis.se/dendro
Article
In 1980 Ernst Hollstein published his Central European Oak Chronology, which covers a period from 724 BCE to 1974 CE. Besides a later correction of the end date of the sampling site chronologies Kirnsulzbach (Germany) and Gustavsburg (Germany) this master chronology has since not been changed and still remains one of the most important bases for dendrochronological dating in western Germany. It stands out in so far as it provides comprehensive graphical findspot series for each individual sampled site and year to year growth values for eight regional sub-chronologies in addition to the combined Central European reference curve. Particularly due to the fact of Hollstein’s chronology being publicly available, it has frequently been criticized for its insufficient data to bridge the Late Antiquity between 350 and 400 CE with only three sampling sites (tomb near Beerlegem, Belgium; tomb inside of Cologne Cathedral, Germany; subfossil trees near Broichweiden, Germany) and that these site chronologies cover those decades with inadequate correlation coefficients. With regard to recent statistical threshold values for crossdating, Hollstein’s Late Antiquity bridging needs to be reconsidered. Therefore, in a combined effort, the dendrochronological laboratories at Rheinisches Landesmuseum Trier (RLM), the University of Cologne and Albert-Ludwigs-University Freiburg re-evaluated Hollstein’s findings for Late Antiquity by including the respective dendrochronological examinations conducted in Rhineland- Palatinate, North Rhine-Westphalia and north-eastern France during the past 40 years. A total of 62 site chronologies were compiled to establish a new Late Antiquity chronology. Thirteen of these site chronologies could be used to support Hollstein’s original bridging series between 350 and 400 CE while the mean series for Broichweiden had to be corrected from end date 365 to 503 CE. Furthermore, this new bridging chronology could be validated by comparing it to an independent chronology from southern Germany. This study thus proves that the integrity of Hollstein’s Central European Oak Chronology is not compromised by a flawed Late Antiquity bridging and that therefore dating based on the Roman part of this chronology can still be considered as absolute.
Verlag Philipp von Zabern, Mainz am Rhein 1980. Digitized mean value site curves download see Appendix B
  • E Mitteleuropäische Hollstein
  • Eichenchronologie
Hollstein, E. Mitteleuropäische Eichenchronologie. Verlag Philipp von Zabern, Mainz am Rhein 1980. Digitized mean value site curves download see Appendix B.
Fällungsdaten Römischer Bauhölzer anhand einer 2350-jährigen Süddeutschen Eichen-Jahrringchronologie
  • B Becker
Becker, B. 1981. Fällungsdaten Römischer Bauhölzer anhand einer 2350-jährigen Süddeutschen Eichen-Jahrringchronologie. Fundberichte aus Baden Württemberg No.6, 369-386.
Paste from a text file" and select your
  • Click
Click "Paste from a text file" and select your.txt-file.
Try to convert data read into a new .fh collection file" and then select a temporary filename
  • Click
Click "Try to convert data read into a new.fh collection file" and then select a temporary filename.