The chronology of culture: a comparative assessment of European Neolithic dating approaches.
Manning, Katie ; Timpson, Adrian ; Colledge, Sue 等
Supplementary material is provided online at
http://antiquity.ac.uk/projgall/manning342/
Introduction
The construction of cultural chronologies in archaeology remains a
point of major ongoing debate. Although the increasing application of
Bayesian analysis has dramatically improved the temporal control that
can be achieved in archaeology by making it possible to integrate
stratigraphic and other information to narrow down calibrated
radiocarbon date distributions (Bayliss & Whittle 2007; Bronk Ramsey
2009; Whittle et al. 2011), such analyses are complex and
time-consuming. At present, stratigraphically constrained Bayesian
analyses of dates are not available in sufficient quantity to provide a
basis for characterising regional and inter-regional patterns in the
cultures of Neolithic Europe, and it is open to question whether this
approach is applicable to such broad scales of archaeological enquiry.
In any case, chronological reasoning, particularly in Neolithic Europe,
is still largely reliant on typological sequences. Although attempts at
providing absolute sequences for cultural chronologies are becoming
increasingly common (e.g. Stadler et al. 2001; Raetzel-Fabian 2002;
Furholt 2003; Reingruber & Thissen 2009; Denaire 2011), there have
been few efforts to systematically compare the beginning and end dates
assigned to cultures, and the date ranges indicated by all of the
available radiocarbon evidence. Some obvious exceptions include, for
example, the review by Wlodarczak (2009) of the Corded Ware dating
methods, and the use of multivariate statistics in conjunction with
radiocarbon data as a means of refining typochronological
classifications. In particular, correspondence analysis and principal
component analysis have been used to identify chronologically meaningful
aspects of regional typologies, e.g. in the Mittelelbe-Saale region in
Germany (Muller et al. 2000; Czebreszuk & Muller 2001; Muller &
Van Willigen 2001, Muller 2009), for Linearbandkeramik (LBK) settlements
in the Netherlands (van de Velde 2012) and also for the Corded Ware
(Ullrich 2008). All of these, however, are relatively localised studies
of internal development.
In this paper, we approach the issue at a different scale, adopting
a broader and therefore lower resolution view of the relationship
between different dating approaches for the European Neolithic; we then
go on to estimate the underlying shape of the intensity of a
culture's presence through time. We investigate the temporal ranges
of various cultural groups by analysing and comparing the estimates from
three sources: the 'standard' date range, the radiocarbon date
range, and the dendrochronological (dendro) date range, where available.
We use the term 'standard' date range to mean the estimated
start and end date for a culture as found in relevant current
literature, which we consider to represent the best-informed
archaeological knowledge on the subject (see online supplementary Table
S1 for all 'standard' dates and the references used). The
motivation behind our analysis came from attempting to investigate
temporal patterns in subsistence data. Given that awareness of available
radiocarbon dates forms an important part of that up-to-date
archaeological knowledge, we were surprised to find a large difference
in our results when using site phases that were dated using associated
radiocarbon samples, compared to the current estimate of the
'standard' date range for that cultural phase. Therefore, in
order to undertake a valid analysis of subsistence trends, or of any
other archaeological phenomena for that matter, we must first
investigate the congruency between the 'standard' date range
and the contextually associated radiocarbon dates for a given culture.
It is worth pointing out at this stage that our interest in the
temporal ranges of different cultural groups is not intended to promote
the notion of archaeological culture as a monolithic and unchanging
entity: cultures as 'bricks'. On the contrary, one of us has
consistently argued the opposite (e.g. Shennan 1978, 1989), following
Clarke's (1968) polythetic definition of archaeological cultures.
We do not need to take the traditional view of cultures as monolithic
blocks to explore their chronological patterns, which may in fact relate
to specific cultural packages. Just as in genetics a contrast is made
between the analysis of gene trees and population trees, both of them
equally valid for different purposes, so it is equally appropriate in
the case of transmitted cultural variation to work at different scales
of resolution for different purposes. Moreover, we do not need to
subscribe to the idea that our classifications 'carve nature at its
joints' in Plato's famous phrase (Phaedrus, 265d-266a; Fowler
1925) to recognise that some categorisations are better than others. One
key criterion is whether variation between the entities being analysed
is significantly greater than that within them (see Shennan et al.
2014), regardless of the processes that resulted in this patterning,
which often remain the focus of dispute (see Furholt 2014 for an
example). In short, while archaeological 'cultures' may not be
useful for investigating local-scale socio-economic dynamics, they have
remained indispensable for characterizing broad-scale spatial and
temporal trends, such as those which prompted this analysis.
Method and dataset
All radiocarbon data were derived from the database collated by the
EUROEVOL project. These data include a lab code, 14 C age and standard
deviation (SD), latitude and longitude. Each [sup.14]C sample was
calibrated via the IntCal09 calibration curve (Heaton et al. 2009). In
order to warrant inclusion in this analysis, the sample also had to have
an associated culture, assigned either by the excavator, laboratory or
post-excavation analyst. The resultant [sup.14]C dataset for this study
therefore consists of 5594 radiocarbon-dated samples from 1784
archaeological site-phases (see Shennan et al. 2013 for the applied
definition of phase), from 71 Neolithic and Early Bronze Age cultures
(see Figure 1 and online supplementary Table S1). In addition to this
[sup.14]C dataset, a second dendrochronological dataset comprising 350
samples were obtained from the Erziehungsdirektion des Kantons Bern
website (Dendro n.d.).
Each [sup.14]C sample can therefore be considered to have two
properties: a radiocarbon calibrated date, and a 'standard'
date, each with its own probability distribution. Clearly there is a
degree of circularity between these two properties, since the cultural
assignment of a sample may be influenced by its radiocarbon date,
especially in the absence of associated diagnostic material; more
importantly, the current 'standard' date ranges will have been
strongly influenced by the radiocarbon dates themselves. Hence our
surprise at the difference in results when using a site's cultural
assignment as the basis for dating it as opposed to the site's own
radiocarbon dates.
Similarly, each dendro sample can be considered to have two
properties: a dendro date and the 'standard' date of the
culture to which it was assigned. Therefore, we can think of these
relationships from the perspective of estimating the date range of a
culture, by comparing its [sup.14]C date range, its 'standard'
date range, and (where available for a handful of cultures) its dendro
date range. Our objective in what follows is thus to establish how much
agreement there is between these date range estimates, and discuss the
implications of those findings, not least for the analysis of other
archaeological patterns. Breunig (1987) had a similar programme,
although at a larger spatial scale, and of course at that time far less
radiocarbon information was available.
Our study area encompasses central and north-western Europe (Figure
1) and covers the Late Mesolithic through to the Early Bronze Age. The
term 'culture' refers here to the archaeological cultures as
single entities, so, for example, the Cortaillod is treated as a whole
rather than being divided into the sub- phases of the Cortaillod i.e.
ancien, classique and tardif. Three of our cultures--Bell Beaker, Corded
Ware and Linearbandkeramik-are multi-regional, 'culture
groups' in Clarke's (1968) terms.
The dataset included 5452 samples from 1748 site-phases, which had
been assigned to only one of these 71 cultures, while 142 samples (2.5%)
from 36 phases had been assigned to two (or in one case three) of these
cultures (see online Table SI for a list of all cultures). This second
type of cultural assignment may have three interpretations: firstly that
the deposits indicate overlap between two different cultures; secondly,
that they belong to a transitional phase between two cultures; and
thirdly that they have an equal probability of belonging to either
cultural range. It is likely that all three interpretations were used on
different occasions; nevertheless, we have adopted the third
interpretation for all cases. This provides a conservative approach with
the effect of erring on the side of slightly greater uncertainty in the
date range.
[FIGURE 1 OMITTED]
First, we investigate the broad-scale relationship between the
radiocarbon date and the 'standard' cultural date for all
samples combined (Figure 2). Both before and after calibration the exact
radiocarbon date of a sample remains elusive. Instead we have a
distribution, which describes the probability of each possible date.
Similarly, the true 'cultural' date could be any date within a
culture's range, which is assumed to be a uniform distribution. We
generate 1000 random realisations, such that for each realisation every
sample has a random point estimate: both a discrete radiocarbon date and
a discrete 'standard' date obtained from their respective
probability distributions. For each realisation the correlation between
the two estimates (using Pearson's R) and the gradient of a
best-fitted linear model were calculated. The linear model used Deming
regression since both variables have an error distribution. By repeating
the random sampling 1000 times we build up a distribution of possible
values, from which confidence intervals (CIs) can be estimated. These
confidence intervals describe the amount of variance caused by the
uncertainty in the probability distributions.
Having established the overall correlation between our date range
estimates, we turned our attention to assessing the relationship between
the different estimates for each culture separately. We selected
individual cultures that met a minimum sampling criterion to ensure good
geographic and temporal representation, such that a minimum of 18 phases
per culture were required, leaving us with the 22 (31%) best-represented
cultures (Table 1), each of which comprised between 18 and 329
site-phases (mean = 64.4), and between 34 and 1030 samples (mean = 195),
and constituted the majority of the data, comprising 4281 [sup.14]C
dates (79%), and 1416 phases (81%).
Four of our selected cultures also have associated dendro dates
(Corded Ware n = 73, Cortaillod n = 66, Horgen n = 66 and Pfyn n = 55),
which generally provide greater precision for an individual sample than
a typical radiocarbon date (mean dendro error = 6.9 years per sample,
compared to mean [sup.14]C error of 74.7 [sup.14]C years).
Dendrochronology can therefore shed further light on the date range of
each culture, so long as comparisons are made within the relevant region
(see below in regard to the Corded Ware Culture in Switzerland).
For each of these latter cultures, the dendro date range is
reported in Table 1 using the earliest and latest date, and are plotted
in Figure 3 with each sample as a vertical red bar with a width equal to
the samples error. Similarly, the 'standard' dates are
reported as a range between earliest and latest, and plotted as a
uniform distribution, represented by a black bar in Figures 3 and 4. The
radiocarbon date ranges are reported in Table 1 as the two- tailed 95%
interval of a summed probability distribution (SPD), while the full
probability distribution (between 8000-0 BC) is plotted in Figures 3 and
4. Each SPD was constructed in three stages: first, raw radiocarbon
dates (mean and SD) were calibrated (see Shennan et al. 2013 for details
of calibration); second, the calibrated probability distributions from
every sample in the same phase were summed and normalised for unity; and
third, these distributions from every phase in the same culture were
summed and normalised for unity. The rationale behind this approach is
that if random sampling is assumed, the law of large numbers predicts
that as the number of samples increases, the sample distribution becomes
increasingly similar to the true distribution. By summing the individual
samples' probability distributions, the new probability
distribution contains the combined knowledge of each sample, equally
weighted, as each sample was initially considered an equally fair
possibility of the date of the event. This applies at the level of both
the phase and the culture.
In order to assess the level of agreement between the radiocarbon
dates and the 'standard' dates, we calculated the proportion
of the SPD mass that fell outside the 'standard' cultural date
range. This value is reported in Table 1 as the proportion that fell
prior to the start of the 'standard' date range (older), the
proportion that fell subsequent to the 'standard' end date
(younger) and the total proportion.
Finally, we assessed how the intensity or 'floruit' of an
individual culture varied through time, and the underlying shape of this
change. We Z-transformed the SPD for each of the selected 22 cultures,
to ensure they were on the same scale (Figure 5). This involved
obtaining the mean date for the culture's summed probabilities and
the standard deviation of this distribution and then expressing the date
values not on the original year scale but as numbers of standard
deviations away from the mean, thus eliminating the effect created by
the fact that some cultures lasted longer than others. The 22
Z-transformed SPDs were weighted by number of phases per culture, then
summed and normalised to unity (Figure 5, blue line), since we can
expect the 'shape' of those cultures with more phases to be
more representative of the true culture shape. Further, for comparison
we summed and normalised the 22 Z-transformed SPDs with an equal
weighting (Figure 5, red line), to provide some indication of the
efficacy of our minimum inclusion criterion of 18 phases per culture.
Results
Overall correlation
Figure 2 illustrates 10 realisations (out of the total 1000
realisations that were computed) for each of our 5594 radiocarbon
samples plotted in grey. Note the 'blocky' nature of the
scatterplot in the horizontal range, due to the 'standard'
date being sampled from each culture's wide uniform probability
distribution. The vertical range, meanwhile, is scattered more evenly
since each radiocarbon date was sampled from its own relatively narrow
probability distribution. Clearly, when all samples and all cultures are
included, there is a strong correlation between the two dating methods.
(R = 0.859, 95% CI = 0.855-0.862), with almost three quarters of the
variance in either estimate explained by the other (R squared = 0.737,
95% CI = 0.730-0.744). The gradient of the best-fitted linear model
plotted in blue is 1.038 (95% CI = 1.030-1.045), which is to be expected
as both the radiocarbon date range and the 'standard' date
range are two different estimates of the same quantity (the true date
range of the culture).
Comparing individual culture date ranges
Having established an overall strong relationship between the
probability distributions of all radiocarbon dates and
'standard' culture dates, the question remains as to whether
certain cultures correlate better with their associated radiocarbon
dates than others. Figure 3 plots the 'standard', radiocarbon
and dendro date ranges for the four cultures with associated dendro
samples, and Figure 4 plots the 'standard' and radiocarbon
date ranges for the remaining 18 selected cultures. Generally, there is
good agreement between the dendro date ranges and the
'standard' date ranges, which is perhaps not surprising due to
the high level of precision on dendro dates and the excellent
preservation of the wetland sites where these cultures are predominantly
found. The exception to this is the Corded Ware, which has a much wider
'standard' range than the dendro range, due to the fact that
the Corded Ware is found over a much wider geographic area than the
typical wetlands required for dendrochronology, and lake shore
settlement ceased c. 2450 cal BC (Billamboz & Koninger 2008).
[FIGURE 2 OMITTED]
In contrast, the radiocarbon distributions are substantially wider
than the 'standard' date ranges in all cases. Table 1 reports
the proportion of the radiocarbon probability mass that exceeds the
'standard' range, revealing a massive spread of values from
14.8% (Pitted Ware) to 75.1% (Veraza) with a mean of 38.7%. Such a broad
range suggests that some cultures may have more reliable chronological
sequences than others. On a more general note, however, it is clear that
the level of agreement between radiocarbon dates and the conventional
date ranges for European Neolithic cultures remains quite poor. Overall
the radiocarbon ranges tend to be skewed towards being younger than the
'standard' date ranges; on average 23% of the radiocarbon
probability mass is younger than the 'standard' date range and
15% is older (22% and 12% respectively when weighted by sample size).
This is rather surprising in light of our inclusion of charcoal dates
that might lead to 'old wood' effects.
Other differences in the nature of this discrepancy for each
culture are evident. Some, for example Veraza and Iwienska, indicate an
overall shift in the peak of radiocarbon probability density, suggesting
the 'standard' range may be slightly too early or too late.
Others, for example Seine-Oise-Marne, have a generally broader
radiocarbon distribution, indicating a potential underestimation of the
'standard' date range. The specifics of some of these cultural
groups are discussed below in relation to the potential underlying
causes of chronological discrepancies.
[FIGURE 3 OMITTED]
The 'temporal shape' of Neolithic cultures
In addition to comparing different dating methods, the technique of
summing probability distributions also allowed us to investigate the
broad-scale 'temporal shape' of Neolithic cultures. Care
should be taken when interpreting the distribution of individual
cultures across relatively narrow ranges since the radiocarbon
calibration process often generates spurious wiggles on the short-term
scale (below around 200 years). By Z-transforming each SPD we can
compare the shape of each culture's radiocarbon distribution on a
like-for-like scale. This transformation also had the benefit of causing
these wiggles to shift and change in size, removing their constructive
interference and instead reducing them to random background noise in the
overall shape.
Figure 5 shows SPDs after Z-score transformation for each of our 22
cultures. The red line shows the mean of all cultures, while the blue
line shows the mean weighted by the number of phases per culture. Both
lines show remarkable similarity, suggesting that our inclusion
criterion of a minimum 18 phases was successful in ensuring each SPD was
well representative of the culture's true underlying distribution.
The dashed black line shows a standard normal distribution (mean = 0, SD
= 1), which describes the shape of any normal distribution after Z-score
transformation. The fit is extremely compelling.
[FIGURE 4 OMITTED]
[FIGURE 5 OMITTED]
The principal advantage of this normal distribution is its
increased explanatory power at no extra cost of parameters. Currently,
the 'standard' date range uses two parameters: a start and end
date, while a normal distribution also only requires two parameters: the
mean and SD. For example, currently the 'standard' date range
for the Linearbandkeramik is described using start = 5500 BC and end =
4900 BC, which tells us little about the actual distribution of the
available evidence. In contrast, a normal distribution more accurately
describes its shape, including its central tendency as well as duration,
using mean = 5088 BC and SD = 310 yrs. Hence, the additional information
that is gained, at no extra cost, allows us to begin exploring the
statistical properties of cultural chronologies in a way that is not
possible with the current start and end date ranges. Figure 6 shows
examples of the fitted normal distribution with its mean and SD for five
selected cultures. The mean and SD are reported for all 22 cultures in
Table 2, and are plotted in Figure S1 in the supplementary information.
Discussion
It may be fair to assume that the additional comparative dates
provided by dendrochronology would improve the overall reliability of a
culture's chronology. Although this does appear to be the case for
the Horgen and Corded Ware (only 17% and 23% of ,4C probability mass
outside their respective 'standard' dates), it does not apply
to all cultures; for example, there is a discordance of 51 % in the Pfyn
data. This is in contrast to cultures such as the Pitted Ware, Chasseen
and Michelsberg, which have no associated dendro dates, and yet show
relatively good agreement between the radiocarbon and
'standard' date ranges (15%, 25% and 26% discordance
respectively). This may be due to recent efforts aimed specifically at
refining cultural chronologies such as the Michelsberg and its
successive cultures (e.g. Raetzel-Fabian 2000, 2002; Geschwinde &
Raetzel-Fabian 2009), or because of the archaeological visibility of
particular cultural traits, e.g. the enclosures of the Michelsberg or
the Chasseen, and the distinctive material and economic characteristics
of the Pitted Ware. In contrast, the cultures with the highest
disagreement values, such as the Veraza, Neolithique moyen II and
SeineOise-Marne, are comparatively ill-defined and with relatively small
sample sizes (38, 96 and 39 respectively).
[FIGURE 6 OMITTED]
These results raise a number of important issues relating not only
to chronological reasoning, but also to the characterisation of
Neolithic cultures. On a very basic level, the amount of agreement
between different dating methods at the broad level of analysis we have
used seems to come down to economy and scale. The more established a
culture is, and to some degree the more widespread it is, the more
agreement there is between estimates, implying greater accuracy.
However, this is not simply attributable to sample size, and there is no
clear correlation between sample size and the agreement between the two
estimates. Trichterbecher has by far the largest [sup.14]C sample size
(1030), yet suffers 31% discordance between estimates, whilst Pitted
Ware has fewer samples by an order of magnitude (130), yet has a smaller
discordance of 15%. The Linearbandkeramik enjoys the second highest
sample size of 535 but fares even worse, with a discordance between
estimates of 39%. Despite the relatively broad geographic spread of the
Cardial culture there is high discordance between estimates (51%). Hence
there is no clear-cut cause for discrepancy between the different date
estimates, and instead each culture needs to be considered individually
in light of its own historical trajectory, environmental setting and
history of research.
A more interesting result to have arisen from this analysis is the
comparative shape of the SPD of each culture's 14C date range
estimate. The start and end dates of the 'standard' range
provide us only with a simple uniform distribution, and while this may
tell us something about the duration of a cultural phenomenon, it tells
us nothing about its temporal nature, i.e. its floruit or intensity
(Ottaway 1973; Aitchison et al. 1991).
It is improbable that the material associated with a given culture
suddenly comes into existence maintaining a constant prevalence and then
suddenly ceases. Our analysis from combining all 22 SPDs suggests that a
normal distribution provides a better description of the fundamental
underlying shape of a culture's rise and fall, and, therefore, the
use of a mean and standard deviation would be more informative than the
current start and end dates, which are so often used in the
archaeological literature. Indeed, in some circumstances it may also be
useful to supplement reporting the mean and SD of a cultural range by
reporting the full [sup.14] probability distribution, in order to
incorporate fine-grained information about the temporal variation in the
intensity of the episode. This requires caution as finer resolution may
introduce undesirable artefacts from sampling error and
'interference' effects from wiggles in the calibration curve
(Blockley et al. 2000). As such, this would only be useful where large
unbiased samples are available and when the culture spans a long time
period. Nevertheless, this study provides empirical support that a
Gaussian model, and perhaps other unimodal distributions (Lee &
Bronk Ramsey 2012), are superior to the standard uniform distribution,
both as a prior for Bayesian models of archaeological chronologies
(Bronk Ramsey 2009) and as a simple summary description of the cultural
time-span.
This improved underlying model of the expected temporal dynamics of
a regional culture is of value for a range of different analyses.
Bayesian analysis of [sup.14] dates aggregated at the level of culture
is reliant on the selection of a prior distribution. In the absence of
other information a uniform prior is usually adopted; instead, a
normally distributed prior is both more reasonable and empirically
supported. Analysis requiring Monte Carlo simulation of cultural dates
(e.g. Crema 2013) would also benefit greatly from this improved model.
Finally, it is worth pointing out that the characteristic normal
distribution we have identified bears a striking resemblance to the
so-called 'battleship curves' produced when frequency
seriations are carried out on individual artefact types that are
chronologically sensitive. When cultures are taken as entities they seem
to mirror this effect. In essence, the number of dated events that
archaeologists are prepared to label, for example as Horgen or
Michelsberg, starts small, increases to a peak and then declines again.
The pattern could arise because of the waxing and waning popularity of
temporally correlated styles across a geographical region. Another
possibility, supported by demographic proxies in some cases (Shennan et
al. 2013), is that they reflect fluctuations in local populations; at
some periods there are simply more people in the region, so the number
of dated events characterised by the styles of the period is also bound
to be greater. Of course, these two possibilities are not mutually
exclusive, and in some cases it could be that new cultural innovations
themselves result in periods of population increase. These questions
remain open for subsequent analysis.
Acknowledgements
We are grateful to all those who provided radiocarbon data, and to
Johannes Miiller and Felix Riede for their constructive review comments.
This research was funded by the European Research Council Advanced Grant
number 249390 to Stephen Shennan for the EUROEVOL Project. Information
on the EUROEVOL Project is available at http://www.ucl.ac.uk/euroevol/
References
AITCHISON, T, B. OTTAWAY & A.S. AL-RUZAIZA. 1991. Summarizing a
group of [sup.14]C dates on the historical time scale: with a worked
example from the Late Neolithic of Bavaria. Antiquity 65: 108-16.
BAYLISS, A.L. & A.W.R. WHITTLE (ed.). 2007. Histories of the
dead: building chronologies for five southern British long barrows
{Cambridge Archaeological Journal 17, Supplement 1). Cambridge:
Cambridge University Press.
BILLAMBOZ, A. & J. KONINGER. 2008. Dendroarchaologische
Untersuchungen zur Besiedlungs- und Landschaftsentwicklung im
Neolithikum des westlichen Bodenseegebietes, in W. Dorfler, J. Muller
& A. Wesse (ed.) Umwelt--Wirtschafi--Siedlungen im dritten
vorchristlichen Jahrtausend Mitteleuropas und Siidskandinaviens: 317-34.
Neumunster: Wachholtz.
BLOCKLEY, S.P.E., R.E. DONAHUE & A.M. POLLARD. 2000.
Radiocarbon calibration and Late Glacial occupation in north-west
Europe. Antiquity 74: 112-21.
BREUNIG, P. 1987. [sup.14]C-Chronologie des Vorderasiatischen,
Suedost- und Mitteleuropaeischen Neolithikums. Koln: Boehlau.
BRONK RAMSEY, C. 2009. Bayesian analysis of radiocarbon dates.
Radiocarbon 51: 337-60.
CLARKE, D.L. 1968. Analytical archaeology. London: Methuen.
CREMA, E. 2013. Cycles of change in Jomon settlement: a case study
from eastern Tokyo Bay. Antiquity 87: 1169-81.
CZEBRESZUK, J. & J. MULLER (ed.). 2001. Die absolute
Chronologie in Mitteleuropa 3000-2000 v. Chr. (Studien zur Archaologie
in Ostmitteleuropa 1). Rahden: Marie Leidorf.
DENAIRE, A. 2011. Chronologie absolue de la sequence
Hinkelstein-Grossgartach-Roessen-Bischheim dans le sud de la plaine du
Rhin superieur et le nord de la Franche-Comte a la lumiere des dernieres
donnees, in A. Denaire, C. Jeunesse & P. Lefranc (ed.) Necropole et
enceintes danubiennes du V millenaire dans le Nord-Est de la France et
le Sud-Ouest de lAllemagne: 9-30. Strasbourg: Maison Interuniversitaire
des Sciences de l'Homme-Alsace, Universite de Strasbourg.
Dendro. n.d. Dendro- und Cl4-Daten der Schweiz. Available at
http://www.erz.be.ch/erz/de/index/
kultur/archaeologie/daten/dendro-_und_c 14daten.html (accessed 29 July
2014).
FOWLER, H.N. (trans.). 1925. Plato in twelve volumes. London:
William Heinemann.
FURHOLT, M. 2003. Die absolutchronologische Datierung der
Schnurkeramik in Mitteleuropa und Siidskandinavien
(Universitatsforschungen zur Prahistorischen Archaologie 101). Bonn: R.
Flabelt.
--2014. Upending a 'totality': re-evaluating Corded Ware
variability in Late Neolithic Europe. Proceedings of the Prehistoric
Society. http://dx.doi.org/10.1017/ppr.2013.20
GESCHWINDE, M. & D. RAETZEL-FABIAN. 2009. EWBSL: Eine
Fallstudie zu den jungneolithischen Erdwerken am Nordrand der
Mittelgebirge (Beitrage zur Archaologie in Niedersachsen 14). Rahden:
Marie Leidorf.
HEATON, T.J., P.G. BLACKWELL & C.E. BUCK. 2009. A Bayesian
approach to the estimation of radiocarbon calibration curves: the
IntCal09 methodology. Radiocarbon 51: 1151-64.
LEE, S. & C. BRONK RAMSEY. 2012. Development and application of
the trapezoidal model for archaeological chronologies. Radiocarbon 54:
107-22. http://dx.doi.org/10.2458/azu_js_rc.v54il. 12397
MULLER, J. 2009. Dating the Neolithic: methodological premises and
absolute chronology. Radiocarbon 51: 721-36.
MULLER, J. & J. VAN WILLIGEN. 2001. New radiocarbon evidence
for European Bell Beakers and the consequences for the diffusion of the
Bell Beaker phenomenon, in F. Nicolis (ed.) Bell Beakers today: 59-80.
Trento: All'Insegna del Giglio.
MULLER, J., C. BECKER, H. BRUCHHAUS, E. KAISER, A. NEUBERT, S.
PICHLER & M. ZABEL. 2000.
Radiokarbonchronologie--Keramiktechnologie-Osteologie-Anthropologie-Raumanalysen. Beitrage zum Neolithikum und zur Fruhbronzezeit im
Mittel-Elbe- Saale-Gebiet. Bericht der Romisch-Germanischen Kommission
80: 25-211.
OTTAWAY, B. 1973. Dispersion diagrams: a new approach to the
display of carbon-14 dates. Archaeometry 15: 5-12.
http://dx.doi.org/10.1111/j.1475-4754.1973.tb00073.x
RAETZEL-FABIAN, D. 2000. Calden.Erdwerk und Bestattungsplatze des
Jungneolithikums. Architektur--Ritual--Chronologie. Bonn: Habelt.
--2002. Absolute chronology and cultural development of the
Wartberg culture in Germany. Available at:
http://www.jungsteinsite.uni-kiel.de/pdf/ 2002_2_fabian.pdf (accessed 29
July 2014).
REINGRUBER, A. & L. THISSEN. 2009. Depending on [sup.14] data:
chronological frameworks in the Neolithic and Chalcolithic of
southeastern Europe. Radiocarbon 51:751-70.
SHENNAN, S.J. 1978. Archaeological cultures': an empirical
investigation, in I. Hodder (ed.) The spatial organisation of culture:
113-39. London: Duckworth.
--1989. Introduction: archaeological approaches to cultural
identity, in S.J. Shennan (ed.) Archaeological approaches to cultural
identity: 1-32. London: Unwin Hyman.
SHENNAN, S., S.S. DOWNEY, A. TIMPSON, K. EDINBOROUGH, S. COLLEDGE,
T. KERIG, K. MANNING & M.G. THOMAS. 2013. Regional population
collapse followed initial agriculture booms in mid-Holocene Europe.
Nature Communications 4: 2486. http://dx.doi.org/10.1038/ncomms3486
SHENNAN, S.J., E. CREMA & T. KERIG. 2014.
Isolation-by-distance, homophily, and core' vs. 'package'
cultural evolution models in Neolithic Europe. Evolution and Human
Behaviour, http://dx.doi.org/10.1016/j.evolhumbehav.2014.09.006
STADLER, E, S. DRAXLER, H. FRIESINGER, W. KUTSCHERA, P. STEIER
&: E.M. Wild. 2001. Absolute chronology for early civilisations in
Austria and Central Europe using [sup.14] dating with accelerator mass
spectrometry with special results for the absolute chronology of Baden
culture, in Studia Danubiana Series Symposia II: Cernavoda III-Boleraz:
541-62. Bucharest: Institutul Roman de Tracologie.
ULLRICH, M. 2008. Endneolithische Siedlungskeramik aus Ergersheim,
Mittelfranken: Untersuchungen zur Chronologie von Schnurkeramik- und
Glockenbechern am Rhein, Main und Neckar. Bonn: Habelt.
VAN DE VELDE, P. 2012. Chronology of the Dutch Neolithic
Bandkeramik culture: a new attempt, in C. Bakels & H. Kamermans
(ed.) The end of our fifth decade (Analecta Praehistorica Leidensia
43/44): 293-305. Leiden: Faculty of Archaeology, Leiden University.
WHITTLE, A.W.R., F.M. HEALY & A. BAYLISS. 2011. Gathering time:
dating the Early Neolithic enclosures of southern Britain and Ireland.
Oxford: Oxbow.
WLODARCZAK, P. 2009. Radiocarbon and dendrochronological dates of
the Corded Ware culture. Radiocarbon 51: 737-49.
Received: 13 December 2013; Accepted: 15 April 2014; Revised: 20
June 2014
Katie Manning (1), * Adrian Timpson (1,2), Sue Colledge (1), Enrico
Crema (1), Kevan Edinborough (1), Tim Kerig (1) & Stephen Shennan
(1)
(1) Institute of Archaeology, University College London, 31-34
Gordon Square, London WC1H 0PY, UK
(2) Research Department of Genetics, Evolution and Environment,
University College London, Darwin Building,
* Gower Street, London WC1E 6BT, UK
Author for correspondence (Email: k.manning@ucl.ac.uk)
Table 1. The 22 best-represented cultures with their start and end
dates estimated using the three different methods, and the mean and
SD using the improved normally distributed model. The last three
columns estimate the difference between 'standard' and radiocarbon
date range.
Radiocarbon Standard
(BC) (BC)
Culture Phases Samples Start End Start End
Bell Beaker 113 228 3040 1607 2500 1800
Western Cardial 57 151 6574 3849 5600 4950
Chasseen 75 213 4663 2952 4400 3400
Corded Ware 189 341 3251 1901 2800 2050
Cortaillod 33 111 4503 2963 4100 3400
Ertebolle 46 360 5709 3478 5450 4100
Globular Amphora 60 145 3411 2009 3100 2700
Horgen 26 130 3540 2525 3500 2750
Iwienska 23 131 2195 1082 2600 1800
Lengyel 39 147 5080 3540 4800 4000
Linearbandkeramik 139 535 5719 4127 5500 4900
Michelsberg 45 126 4529 3158 4250 3500
Neolithique final 42 98 3473 2040 2900 2150
Neolithique moyen II 24 96 4534 2973 3950 3500
Peu Richard 18 56 3488 2483 3300 2900
Pfyn 19 92 4251 3138 3850 3400
Pitted Ware 51 130 3637 2257 3400 2350
Rossen 20 50 5018 4045 4750 4400
Seine-Oise-Marne 26 39 3730 1777 3500 2700
Trichterbecher 329 1030 4323 1793 4100 3000
Veraza 20 38 3622 1989 3400 2850
Wartberg 22 34 3488 2205 3500 2800
Proportion of radiocarbon
Dendrochronology probability mass outside
(BC) the standard range
Culture Start End Older Younger Total
Bell Beaker 0.264 0.046 0.310
Western Cardial 0.157 0.349 0.506
Chasseen 0.134 0.115 0.249
Corded Ware 2751 2418 0.171 0.062 0.233
Cortaillod 3895 3517 0.201 0.147 0.348
Ertebolle 0.077 0.206 0.283
Globular Amphora 0.138 0.363 0.501
Horgen 3482 2765 0.038 0.131 0.169
Iwienska 0.000 0.613 0.613
Lengyel 0.115 0.183 0.298
Linearbandkeramik 0.072 0.313 0.385
Michelsberg 0.150 0.106 0.256
Neolithique final 0.228 0.042 0.270
Neolithique moyen II 0.556 0.083 0.639
Peu Richard 0.173 0.268 0.441
Pfyn 3919 3507 0.359 0.151 0.510
Pitted Ware 0.100 0.048 0.148
Rossen 0.169 0.178 0.347
Seine-Oise-Marne 0.130 0.495 0.625
Trichterbecher 0.049 0.261 0.310
Veraza 0.069 0.682 0.751
Wartberg 0.014 0.302 0.316
Table 2. The proposed mean and SD based
on a normal distribution for all 22 cultures.
Normal
distribution (BC)
Culture Mean SD
Bell Beaker 2313 261
Western Cardial 5126 448
Chasseen 3998 419
Corded Ware 2513 281
Cortaillod 3830 351
Ertebolle 4583 624
Globular Amphora 2806 282
Horgen 3066 276
Iwienska 1681 340
Lengyel 4499 362
Linearbandkeramik 5088 310
Michelsberg 3911 328
Neolithique final 2675 280
Neolithique moyen II 4031 317
Peu Richard 3097 268
Pfyn 3779 265
Pitted Ware 2911 362
Rossen 4620 156
Seine-Oise-Marne 2702 669
Trichterbecher 3432 481
Veraza 2625 426
Wartberg 2968 322
