ArticlePDF Available




Sixty-two 14 C dates are analyzed in combination with a recently established local floating tree-ring sequence for the Early Neolithic site of La Draga (Banyoles, northeast Iberian Peninsula). Archaeological data, radiometric and dendrochronological dates, as well as sedimentary and micro-stratigraphical information are used to build a Bayesian chronological model, using the ChronoModel 2.0 and OxCal 4.4 computer programs, and IntCal 2020 calibration curve. The dendrochronological sequence is analyzed, and partially fixed to the calendrical scale using a wiggle-matching approach. Depositional events and the general stratigraphic sequence are expressed in expanded Harris Matrix diagrams and ordered in a temporal sequence using Allen Algebra. Post-depositional processes affecting the stratigraphic sequence are related both to the phreatic water level and the contemporaneous lakeshore. The most probable chronological model suggests two main Neolithic occupations, that can be divided into no less than three different "phases," including the construction, use and repair of the foundational wooden platforms, as well as evidence for later constructions after the reorganization of the ground surface using travertine slabs. The chronological model is discussed considering both the modern debate on the Climatic oscillations during the period 8000-4800 cal BC, and the origins of the Early Neolithic in the western Mediterranean region.
Radiocarbon, Vol 00, Nr 00, 2022, p 142 DOI:10.1017/RDC.2022.56
© The Author(s), 2022. Published by Cambridge University Press for the Arizona Board of Regents on
behalf of the University of Arizona. This is an Open Access article, distributed under the terms of the
Creative Commons Attribution licence (, which permits
unrestricted re-use, distribution, and reproduction in any medium, provided the original work is
properly cited
V Andreaki1*J A Barcel´o1F Antolín2
3P Gassmann4I Hajdas5
OL´opez-Bult´o1H Martínez-Grau2N Morera1A Palomo6R Piqu´
e1J Revelles7
R Rosillo8X Terradas9
1Universitat Autònoma de Barcelona, Spain
2IPNA, Universidad de Basilea, Switzerland
3Department of Natural Sciences, German Archaeological Institute, Germany
4Dendrocron´ologist, retired from Laboratoire de dendrochronologie de lOffice du patrimoine et de larch´
eologie de
Neuchâtel, Lat´
enium, Switzerland
5Ion Beam Physics, ETH Zürich, Switzerland
6Archaeological Museum of Catalonia, Barcelona, Spain
7IPHES. Universitat Rovira i Virgili. Tarragona, Spain
8Independent researcher, Spain
9Spanish National Research Council (CSIC-IMF), Barcelona, Spain
ABSTRACT.Sixty-two 14C dates are analyzed in combination with a recently established local floating tree-ring
sequence for the Early Neolithic site of La Draga (Banyoles, northeast Iberian Peninsula). Archaeological data,
radiometric and dendrochronological dates, as well as sedimentary and micro-stratigraphical information are used
to build a Bayesian chronological model, using the ChronoModel 2.0 and OxCal 4.4 computer programs, and
IntCal 2020 calibration curve. The dendrochronological sequence is analyzed, and partially fixed to the calendrical
scale using a wiggle-matching approach. Depositional events and the general stratigraphic sequence are expressed
in expanded Harris Matrix diagrams and ordered in a temporal sequence using Allen Algebra. Post-depositional
processes affecting the stratigraphic sequence are related both to the phreatic water level and the contemporaneous
lakeshore. The most probable chronological model suggests two main Neolithic occupations, that can be divided
into no less than three different phases,including the construction, use and repair of the foundational wooden
platforms, as well as evidence for later constructions after the reorganization of the ground surface using travertine
slabs. The chronological model is discussed considering both the modern debate on the Climatic oscillations during
the period 80004800 cal BC, and the origins of the Early Neolithic in the western Mediterranean region.
KEYWORDS: Bayesian analysis, dendrochronology, lakeside settlement, Neolithic, stratigraphy.
The La Draga archaeological site is located on the eastern shore of the Lake of Banyoles, at 172
meters above sea level, in the northeastern part of the Iberian Peninsula. The site is located at
an intermediate point between the Pyrenean Mountain ranges, 4050 km, and 35 km from the
current Mediterranean coastline (Figure 1). Nowadays the site is partially on dry land and
partially covered by the lake water table, and these conditions have favored the
extraordinary state of conservation of built structures and objects made from wood and
vegetable fibers, as well as other organic materials.
Since the sites discovery in 1990, archaeological excavations have documented various structures
that would correspond to an Early Neolithic settlement in which evidence of Cardial pottery has
been identified (Bosch et al. 2000,2006,2011; Tarrús 2008;Palomoetal.2014; Bogdanovic et al.
2015; Terradas et al. 2020). The sites location corresponds to a repeated pattern during the first
*Corresponding author. Email: Published online by Cambridge University Press
Neolithic occupations of the western Mediterranean. These are humid areas, on the shores of lakes,
lagoons, or marshes, and close to land potentially suitable for agricultural practices, in areas of
great ecological diversity (Bernabeu et al. 2017; Guilaine 2018;Revellesetal.2018; Martínez-
Grau et al. 2020;Piqu
eetal.2021). Archaeological studies suggest that the prehistoric
settlement covered an area greater than 15,000 m2. Topographically, the settlement lies on a
smooth downward slope from east to west and from south to north, towards the lakeshore.
So far, a total extension of 1000 m2has been excavated, distributed into three sectors (A, BD
and C) (Figure 1). The first excavations were carried out between 1991 and 2005 (Bosch et al.
2000,2006,2011) discovering an area of 328 m2in Sector A, and 132 m2in Sector B. In the
underwater sector (Sector C), exposed in prehistoric times, 310 m2were also excavated (Bosch
et al. 2000). Archaeological excavations resumed in 2010. A new excavation area of 55.5 m2
was opened, adjacent to Sector B, which was named Sector D, and another new area of 178 m2,
adjacent to Sector A, was also investigated (Palomo et al. 2014).
The oldest well-documented human occupation at La Draga is characterized by the
construction of wooden platforms on piles driven into the lake marl substrate, on the shore
of the lake, above the water table of that time. On top of these wooden platforms, the
dwellings were built, probably with a gable roof (Figure 2). After the abandonment of this
first settlement, a new pavement was built to insulate the surface from the phreatic level,
and a new habitation took place over the same area and sectors.
The entire settlement raised above the lake water table in prehistoric times. In Sectors BD and
C, the oldest stratigraphic layers are found below the water table, which has favored the
preservation of organic materials. In contrast, in Sector A, archaeological layers remained
above the water table for most of their post-depositional history and plant remains have
Figure 1 La Draga archaeological site in northeastern Iberian Peninsula and general plan of the excavated sectors.
The modern lake boundary clearly marks the location of Sector C under the water table.
2V Andreaki et al. Published online by Cambridge University Press
only been preserved by carbonization. The different states of preservation of the archaeological
record at the three excavated areas of the site and the distinct post-depositional processes
generate additional problems when correlating the structures of the different sectors. This is
not only due to the sites elevation in relation to the groundwater, but also the dynamics of
the phreatic level (Figure 3).
Stratigraphic observation suggests the existence of a minimum of two different site
occupations. They are clearly differentiated at Sector BD, as the travertine pavement
overlaps the wooden layer. In Sector A, stratigraphical dynamics are totally different, and
the correlations of the layers and phases are unclear (Palomo et al. 2014). The goal of this
paper is to place both Neolithic occupations in the calendrical scale by integrating
radiometric and dendrochronological dates, as well as sedimentary and micro-stratigraphic
Figure 2 Virtual and idealized reconstruction of one of the Neolithic huts that may have been built at La Draga
during its early occupation (Campana 2019; Barcel´o et al. 2020). Pile alignment in the geometrical model does not
reflect the reality of the terrain.
Figure 3 Stratigraphic correlation of all three excavated sectors in La Draga. (Modified after Palomo et al. 2017.)
Absolute Chronology at the Waterlogged Site of La Draga 3 Published online by Cambridge University Press
information and archaeological materials. The resulting chronological model is discussed amid
current debates on the origins of the Early Neolithic in the western Mediterranean region and
the Cardial pottery context.
Methodology and Theoretical Background
All radiocarbon samples and tree-ring data have been organized into depositional events, using
archaeological contexts and micro-stratigraphic information. In so doing, we have followed the
general approach by Barcel´o and Andreaki (2020) and Barcel´o and Bogdanovic (2020).
Adepositional event is the material expression of an archaeological event: something
happened at a specific place during an interval of time and modified the physical
appearance of the ground surface where the action took place, differentiating specific areas
from their neighbors. Therefore, a depositional event will correspond to the smallest
differentiable spatial unit, that is, a particular closed area where the values of some spatial
variable(s) are homogenous and statistically different from the values the same spatial
variable(s) had at neighboring closed areas. It is important to consider that social action
alone is not the only cause for a depositional event to occur. This is because an
archaeological site is not only the place where human action took place at a certain time,
but also, where numerous post-depositional processes (geological, chemical, physical,
mechanical, biological, etc.) modified or altered that initial anthropic deposition.
A depositional event is thus the smallest spatial referential archaeological unit of observation
showing some degree of homogeneity, and it should be defined according to the modification of
the surface generated by the activity at that place: the accumulation of materials on the surface,
and/or the excavation of the same ground surface. In so doing, depositional events should be
defined according to the following:
the archaeological materials they contain (what has been deposited),
the microstratigraphical information revealing the formation processes in situ (the way the
ground was altered as a result of deposition),
the relative (stratigraphic order) and absolute chronological information
(dendrochronological and radiocarbon data) for each of them (the order in which
different depositions occurred, and the position of each deposition in the calendar scale)
(Barcel´o and Andreaki 2020).
It is generally assumed that the most likely position of a depositional event on the time scale should
be close to the temporal position of a majority of the isotopic events it contains (Barcel´oand
Bogdanovic 2020). Van der Plicht et al. (1999: 434) referred to 14C events,definingthemas
the separation of a certain substance containing carbon from the source from which the
carbon was obtained.Generalizing to any kind of isotopic clock, we refer to an isotopic event
(see also, Lanos and Philippe 2017,2018,2020). Since the temporal position of each isotopic
event on the time scale is necessarily uncertain, we need a series of isotopic events (the larger,
the better), and a combination of their respective confidence intervals, to estimate the temporal
position of the depositional event. The rationale for this chronological inference is that the
material consequences of activities performed at the same time should be closer spatially to
each other than to those materials that were deposited farthest in time. Synchronicity of
depositional processes suggests that all things being equal, activities occurring at the same time
4V Andreaki et al. Published online by Cambridge University Press
will tend to increase the joint frequency of their effects, and this can be observed in the spatial
density of such effects. Nevertheless, this assumption is not always correct. Not only must we
keep in mind the possible time lag between the isotopic event and the depositional event, but
we must also know in detail the content-container relationship that may exist between the
dated sample and the minimum spatial unit of reference in which it has been found (Roskams
1992;Berry2008; Thorpe 2012). To solve the question whether all the materials found at a
referential spatial unit were deposited at the same time, the significance of the differences in the
estimated 14C age can be calculated using the classical Ward and Wilson test (1978). If the
result is positive for all the isotopic events within the same depositional event, we can conclude
that the duration of the deposition was short, and the position in the calendar scale of the
depositional event will be calculated in terms of the statistical combination (average) of the
uncalibrated 14C ages of the samples contained in the spatial unit.
In the case that the combination of isotopic events from the same depositional event fails the
Ward and Wilson statistical test, the estimation of the temporal position of the depositional
event will be compromised. The samples either were deposited as a consequence of different
depositional actions, or the time lag between different effects of the same action is too
great to be effectively detected in terms of a chi-square statistical distribution. Depositional
events can be fasta day, a week, less than a year; medium slowless than 20 years; or
slowmore than 20 years. That is the reason why some authors suggest distinguishing
between strict contemporaneityand broad contemporaneity(Sharon 1995; Holst 2001;
Desachy 2008). Two elements are strictly contemporaneous if they were deposited at
exactly the same time; broadcontemporaneity expresses those two depositional events
that may have occurred within the same temporal interval, but not necessarily
simultaneously. The duration of the depositional process, its continuity, and the longitude
of the temporal gap between the start and the end of the deposition also introduce strong
and weak synchronisms. This distinction is important, because weak synchronisms or broad
contemporaneitieswould not allow us to establish temporally ordered relationships: if A
is later than B, and if Bis possibly contemporaneous with C(but we do not know with
certainty), then we cannot affirm that Ais later than C, although the degree of certainty
can be expressed in probabilistic terms.
The slower the depositional event, and the longer time it needed to end, the more difficult it can
be to fix the temporal positioning on the calendar scale. On one hand, the statistical
combination of estimated dates for isotopic events in the same depositional event is more
difficult, because the precision of radiocarbon estimate is often lostin the calibration to
calendar time scale (Blaauw et al. 2005). In the scenario the depositional event is assumed
to have occurred along a relatively large time span (it was slow), the contemporaneity of
constituting isotopic events is broad,or the different depositional events seem to be
functionally related in some way, we can build chronological units defined as phases.
We have used OxCal 4.4. (Bronk Ramsey 1994,2019) and ChronoModel 2.0.18 (Lanos et al.
2016; Lanos and Dufresne 2019) software tools to integrate isotopic events that were
depositionally associated with a single depositional event. In addition to that, depositional
events were ordered according to stratigraphic constraints. The term phaseis used here
in the same way as Bronk Ramsey (2015) and Lanos and Philippe (2018,2020): as a group
of eventsisotopic and/or depositionalthat are related in some way but for which there
is no information on the internal ordering, and no (prior) chronological distinctions or
temporal ordering can be assumed. That means, that phases are undetermined temporal
Absolute Chronology at the Waterlogged Site of La Draga 5 Published online by Cambridge University Press
intervals and the only way of estimating their temporal position depends on the probability of
fixing the temporality of their start and end events (Andreaki et al. 2020).
In the remaining of the paper, a phase is thus an aggregation of broadly contemporary
depositional events, or a single depositional event whose formation process has been
extremely slow. Martín-Rodilla et al. (2016) qualify the supposed contemporaneity of
events belonging to the same phase by saying that it is a circumstance occurring over a
long-time interval, during which no changes appear in the associated entities. It is the
interval of the calendar scale fulfilling the condition: there is a non-zero and calculable
probability that any depositional event included within its limits contains at least one of the
isotopic events to which it refers(Barcel´o 2009). Furthermore, phases are groups of
functionally linked archaeological units, in the sense expressed by E. Harris (1989): they are
the result of a structural combination of structural archaeological spatial reference units,
and not necessarily of temporal (chronostratigraphic) units (see also Cox 2001; Traxler and
Neubauer 2008). Although different, both uses of the term phasehave similar
explanations when used to reconstruct the biographyof an archaeological site. They can
be viewed as individual steps in the temporal trajectory of the site occupation and
formation. In both cases, a single scalar calendar date for positioning such steps is not
enough, however, we could fix in some way the start and end of the activity or processes
responsible for the formation of the individual event or the functionally connected set of events.
Stratigraphy and the Ordering of Depositional Events
The Harris Matrix diagrams (Figure 4) describing stratigraphic relationships between
excavated units were translated into a sequence of depositional events, whose temporal
Figure 4 Harris Matrix Diagrams from Sectors D (left) and A (right) at La Draga. The stratigraphic units are
defined either as surfaces (green) or deposits (blue) and are organized in three phases in Sector D and four
phases in Sector A.
6V Andreaki et al. Published online by Cambridge University Press
range depend on the isotopic events depositionally identified at each minimum spatial unit of
The oldest depositional events correspond to the construction of foundational wooden
platforms, and integrate the isotopic events measured from the foundational vertical piles
at each sector: Event 1 in Sector A, Event 2 in Sector B, and Event 3 in Sector D.
In Sector A, more than 400 wooden pile tips have been recovered stuck deep in the geological
lake marl under the archaeological level. Only the bottom part of the vertical piles driven into
the original carbonate sands has been preserved (Event 1), while the occupation layers with
organic and anthropogenic remains are affected and compacted by direct or indirect
trampling from the travertine slabs above. Event 4 corresponds to the piles used for the
successive repair of the platforms documented at this area, but also archaeological material
found in the sedimentary filling of post-holes.
Sectors B and D, where wooden elements and organic material have been very well preserved
under the actual phreatic level, present a more complex stratigraphy. High-resolution
microstratigraphic analysis (Andreaki 2022) reveals a compact sediment, because of
trampling action of the surface, while at the same time the decomposition of organic matter
is observed. Sector Bis the closest to the old lake shoreline. Radiocarbon dated samples
from wooden piles related with platform repair have been integrated into Depositional
Event 5. Event 8 includes nine faunal bones, cereal seeds and wooden tools found in
contact with preserved wooden elements. A further distinction has been made between
materials found directly on the occupation surface (Depositional Event 8a) and organic
material over it but in close connection with the collapsed wooden remains (Depositional
Event 8b).
Over the lake marl surface, a dark organic sediment, NAVIII, is found in some parts of Sector
D; several well-preserved leaves have been identified in this layer. A dark grey sediment,
NAVII, with abundant and well-preserved wooden elements (tools, branches, twigs, boards)
and other organic remains as leaves or fungi, associated with charcoals, and accumulations
of cereal seeds, stratigraphic unit 7001, was found in between the wooden elements.
Depositional Event 6a includes a single associated isotopic event, a cereal seed (Beta
315052), found at the bottom of stratigraphic unit NAVII. Event 6b is defined by the
presence of four additional cereal seeds (Hordeum and Triticum), associated with domestic
activities that occurred on the wooden platforms (filling of stratigraphic unit NAVII), and
probably also affected by minor post-depositional activity related with the fluctuations of
the water table. Depositional Event 28 is also functionally associated with Event 6b, and it
corresponds to stratigraphic unit 7001 from which, two seeds (Triticum and Papaver
somniferum) were dated.
In Sector C, currently submerged underwater, the stratigraphic sequence is affected by
subsequent lake marl depositions, preceding and following the archaeological layers. The
alteration between depositional processes of peat, carbonate sandy sediments and lake marl
silt is usual in wetland sites. The first archaeological layer, in close contact with the
carbonated sands of the original lake ground, has a mean thickness of 15 cm. It is
characterized by the presence of wooden elements, and a big amount of vegetal remains, as
well as remains of fauna, pottery, and animal bones (Depositional Events 7a and 7b).
Absolute Chronology at the Waterlogged Site of La Draga 7 Published online by Cambridge University Press
At some point, after the construction of wooden platforms and the building of dwellings and
other functional structures on them, the ground surface subsided and as a result, lake water
partially flooded parts of the settlement. Current phreatic level can be related to the lake
water level fluctuations. At the same time, a ground subsidence of the original lake marl
substrate affecting Sectors B and D of the current site has been geomorphologically
identified (Iriarte et al. 2014), and it may have contributed to the collapse of the old
wooden structures in this part of the site. The interplay of water input and sediment
accumulation rate constrained the continuity of human activity.
The practical response to ground sinking and flooding was probably to insulate the swamped
and partially inundated surface with locally available travertine slabs in Sector BD, as the
ground subsidence continued to grow with the passage of time. In this sector, immediately
below this accumulation of travertine slabs, the sediment appears to be ashy and oxidated
towards the top making more probable the hypothesis of an insulation layer before the
deposition of travertine as the influence of dry conditions is greater. Over the travertine
layer, a peaty layer was gradually formed with charred plant material, faunal remains and
In Sector A, most travertine slabs marking the probable second occupational surface are in
contact with the original lake marl surface, probably because of the poor preservation of
the wooden platforms at this sector of the site. Depositional Event 9 can be defined based
on a radiocarbon dated sample from stratigraphic layer IIIB, in contact with the travertine
slabs. Forty combustion features (hearths) have been identified, arranged with travertine
slabs, sandstones or burnt pebbles, and include charcoal, remains of the firewood used and
other burnt and unburnt material. They appear in the form of pits with basal depression of
8090 cm in length, and with a sedimentary filling of 1020 cm thick, approximately. The
stratigraphical sequence of most of those hearths is very characteristic: a first layer
containing some charcoals and mixed archaeological material, fragmented travertine slabs
and some sandstones and a second layer above, with a bigger number of charcoals. The
top of the second layer is covered by a new accumulation of travertine slabs. Depositional
Events 1019 have been identified corresponding to samples dated from different hearths.
Cereal seeds and charcoal remains in 10 out of the 40 differentiated hearths contribute to
their chronological definition.
Apart from the identified hearths, there are other differentiated spatial units in Sector A, also
formed by arranged travertine slabs of various measures, with a certain basal depression
approaching 20 cm of thickness. These distinct structures are filled with large quantities of
diverse archaeological material such as charred seeds, animal bones, fragments of pottery,
quartz, flint and bone tools, pieces of ornaments and grinding tools. Because of the kind of
materials they contain, such structures have been interpreted as landfills for food waste and
remains of discarded objects (rubbish) (Figure 5).
Depositional Event 20 corresponds to E254, an irregularly shaped pit filled with a brownish
grey sediment of clayey texture (total extension: 5.20 ×3.80 m) and containing abundant
archaeological material such ornamental objects. A single isotopic event corresponds to it.
Event 21 defines the temporal position of E260, another small oval shaped arrangement of
travertine slabs (75 ×54 cm), with a maximum thickness of 17 cm, found over the lake
marl substrate. It is filled with a greyish clayey sediment and contains travertine slabs of
different measurements between 5 and 30 cm. The material recovered, apart from the slabs,
8V Andreaki et al. Published online by Cambridge University Press
is characterized by an accumulation of faunal remains anatomically connected. An isotopic
event associated to this deposition has been measured from a bone.
Depositional Event 24 corresponds to the formation of E263, another arrangement of
travertine slabs, whose bottom part was dug into the lake marl substrate. Its sediment is
organic of darkish color, containing charcoals and a large amount of archaeologic material,
especially faunal remains. A single isotopic event from this deposit comes from an animal
bone fragment. An isolated seed coming from a concentration of pottery sherds located in
an extreme corner of the excavated area contributes to defining Event 27.
E258, an irregular oval shaped arrangement (5.10 ×1.30 m) of different sized travertine slabs
between 5 and 50 cm, has been divided into two differentiated depositional events (Events 22
and 25). Although the structure is filled with a homogenous dark clayey sediment containing a
big number of charcoals, sedimentary differentiation between the bottom and the top allows
distinguishing two different moments in its construction and filling. A single isotopic event for
each of these depositional events comes from animal bones identified at a precise location. This
is also the case of E261, a big distinctive spatial unit (5.40 ×2.95 m), with a basal depression
dug into the lake marl substrate and a filling sediment and content like E258. Its differentiated
sedimentary sublayers have been distinguished depositionally (Events 23 and 26).
Stratigraphic and depositional units defined in Sector A can be explained as synchronous and/
or post-depositions of travertine arranged features. A darkish brown sediment with some
modern archaeological material, extending all over Sector A, covers them. The top of the
Figure 5 Spatial distribution of the excavated structures in Sector A mentioned in this analysis, La Draga (Morera
and Terradas 2017).
Absolute Chronology at the Waterlogged Site of La Draga 9 Published online by Cambridge University Press
second layer is covered by a new accumulation of travertine slabs. Overlying the travertine
associated features, there is a darkish grey sediment, marking the end of travertine use in
Sector A.
The second occupation in Sector D seems to be slightly different. It contains stratigraphic units
characterized as pre-depositional and/or synchronous with the paved surface made of
travertine slabs of various sizes. This includes, stratigraphic unit NAVI extending all over
the excavated sector immediately below the paved surface of travertine slabs and just above
the preserved wooden elements from previous occupation. Stratigraphic units NAVIa,
NAVIa-a, and 6005 have been detected only in some parts of this sector. Stratigraphic unit
NAV corresponds to travertine slabs defining an apparently paved surface, and NAIV is
found above this accumulation of travertine slabs. These sediments are of terrestrial origin,
either from fluvial transport or from accumulation after torrential rains, and they deposited
in a very short time interval covering the trenches and basins caused by ground surface
subsidence. Those layers are mainly composed of clays, and very poor in archaeological
The fact that the plant remains from this second occupation have only been preserved by
carbonization suggests that these more recent layers remained above the water table for
most of their post-depositional history. Event 29 contains a cereal grain coming from a
stratigraphical layer immediately above the wooden collapse layer. It is defined
sedimentologically by clays with plastic texture and a high presence of organic material,
and extends throughout the excavated Sector D. Stratigraphically above it, depositional
event 30 is defined by another cereal grain coming from a restricted area characterized by a
peaty sediment of dark color, with little presence of archaeological material and mostly
consisting of decayed organic matter, whose inferior part is in contact with the travertine
paved area.
Apart from these depositional events, additional excavated units have been associated to
second occupation syn-depositional and/or post-depositional events. These are stratigraphic
units III, II and I. Stratigraphic unit III contains fragmented travertine slabs and coincides
with the upper part of the paved surface. Stratigraphic units II and Iare clearly post-
depositional events, as they consist of a darkish sediment with decayed organic matter and
scarce mixed archaeological material. Event 31 contains a sample of fauna from a sandy
greyish sediment containing travertine sand and fragmented slabs and is stratigraphically
correlated with this upper part of the travertine slabs found in the paved area.
Radiocarbon Dating
There are 62 14C dates from short and medium long-lived samples from all the sectors of the site
(Table 1). Cereal and fauna samples refer to domesticated species. All dates correspond to
singular elements, and their precise 3D location and stratigraphical association allows
assigning all of them to depositional events.
The dated samples correspond mainly to short-lived samples as cereal seeds and bones, but also
charcoal and wooden material retrieved during the excavation, among them several wooden
piles. In order to avoid the old wood effect, when possible, the last rings of the wooden piles
were sampled for radiocarbon dating. Sectors A and BD are the best dated, with 29 dated
samples coming from Sector A, 15 from Sector B, and 12 from Sector D. In contrast, we
dispose of only six radiocarbon dated samples from the underwater Sector C.
10 V Andreaki et al. Published online by Cambridge University Press
Table 1 Radiocarbon (14C) dates from archaeological samples in La Draga and their respective ChronoModel events. δ13C and δ15N values
of the samples are also included.
Lab ID Sample Context Sector Method
Ua-62940 Quercus sp.
Wooden post PV089 A AMS 6401 38 26.8 Event 1. Construction
Beta-453513 Laurus nobilis Wooden post PV1300,
Structure 261
A AMS 6280 30 28.5 Event 1. Construction
Beta-481571 Quercus sp.
Wooden post PV1311,
Structure 260
A AMS 6270 30 25.08 Event 1. Construction
UBAR-314 Quercus sp.
Wooden post PV106 A CON 6410 70 Event 1. Construction
Beta-425194 Quercus sp.
Wooden post PV1399,
Structure E258
A AMS 6170 30 26.8 Event 1. Construction
Ua-62941 Quercus sp.
Wooden post PV738 B AMS 6308 39 27.8 Event 2. Construction
UBAR-1308 Quercus sp.
Wooden post PV605 B CON 6270 45 26.77 Event 2. Construction
Ua-62942 Quercus sp.
Wooden post PV986 D AMS 6285 39 27.1 Event 3. Construction
Beta-425196 Quercus sp.
Wooden post PV153,
Structure E73
A AMS 6310 30 25.7 Event 4. Repair
Beta-481572 Quercus sp.
Wooden post PV1441,
Structure E263
A AMS 6320 30 25.93 Event 4. Repair
Beta-425195 Quercus sp.
Wooden post PV191,
Structure E6
A AMS 6260 30 26.5 Event 4. Repair
1Years BP refer to conventional radiocarbon ages (present is AD1950) (Stuiver and Polach 1977).
Absolute Chronology at the Waterlogged Site of La Draga 11 Published online by Cambridge University Press
Table 1 (Continued )
Lab ID Sample Context Sector Method
Beta-505910 Quercus sp.
Wooden post PV1450 A AMS 6210 30 27.2 Event 4. Repair
Beta-453512 Charcoal, Quercus
sp. deciduous
Structure E263 A AMS 6280 30 25.4 Event 4. Use
UBAR-1247 Quercus sp.
Wooden post PV582 B CON 6295 45 27.19 Event 5. Repair
UBAR-1248 Quercus sp.
Wooden post PV584 B CON 6240 35 25.08 Event 5. Repair
UBAR-1293 Wood Wooden post PV600 B CON 6220 45 28.19 Event 5. Repair
UBAR-1309 Wood Wooden post PV607 B CON 6205 45 27.46 Event 5. Repair
Beta-315052 Cereal Layer VII D AMS 6310 30 22.7 Event 6a. Use
ETH-88874 Hordeum vulgare Layer VII D AMS 6152 26 23.4 Event 6b. Use
ETH-88873 Triticum aestivum/
Layer VII D AMS 6131 26 24.5 Event 6b. Use
Beta-315049 Cereal Layer VII D AMS 6140 40 24.5 Event 6b. Use
ETH-88872 Triticum aestivum/
Layer VII D AMS 6116 26 25.0 Event 6b. Use
ETH-88875 Triticum aestivum/
Layer E7001 D AMS 6110 26 25.0 Event 28. Use
Echo-2453.1.1 Papaver somniferum Layer E7001 D AMS 6060 110 Event 28. Use
Beta-278255 Fauna Underwater layer II C CON 6270 40 21.4 Event 7a. Use
Beta-278256 Fauna Underwater layer II C CON 6170 40 21.1 Event 7a. Use
ETH-88870 Cereal Underwater layer II C AMS 6098 26 24.7 Event 7b. Use
ETH-88871 Cereal Underwater layer II C AMS 6123 26 24.8 Event 7b. Use
Beta-137197 Quercus sp.
Wooden tool, Layer II B AMS 6290 70 25.0 Event 8a. Use
12 V Andreaki et al. Published online by Cambridge University Press
Table 1 (Continued )
Lab ID Sample Context Sector Method
Beta-137198 Buxus sempervirens Wooden tool, Layer II B AMS 6270 70 25.0 Event 8a. Use
Beta-588213 Fauna Layer II B AMS 6260 30 21.0 3.9 Event 8a. Use
Beta-00002Fauna Layer II B CON 6184 27 Event 8b. Use
OxA-20231 Cereal Layer II B AMS 6163 31 23.4 Event 8b. Use
OxA-20232 Cereal Layer II B AMS 6121 33 23.4 Event 8b. Use
Echo-2448.1.1 Papaver somniferum Layer II B AMS 6090 90 Event 8b. Use
ETH-88869 Cereal Layer II B AMS 6142 26 25.6 Event 8b. Use
Beta-588214 Fauna Layer II B AMS 6100 30 21.2 6.5 Event 8b. Use
OxA-20233 Cereal Layer IIIb A AMS 6179 33 22.3 Event 9. Second
OxA-20235 Cereal Structure E21 A AMS 6143 33 22.7 Event 10. Second
Beta-438952 Triticum durum Structure E6 A AMS 6150 30 24.3 Event 11. Second
OxA-20234 Cereal Structure E5 A AMS 6127 33 22.5 Event 12. Second
HD-15451 Cereal Structure E3 A AMS 6060 40 Event 13. Second
UBAR-313 Cereal Structure E56 A CON 6010 70 Event 14. Second
ETH-88867 Cereal Structure E14 A AMS 6108 26 24.5 Event 15. Second
2Beta-0000 is a date of which we dispose no lab number at the moment. However, the 14C date has been already used in previous publications (Colominas et al. 2015; Bogdanovic
et al. 2015).
Absolute Chronology at the Waterlogged Site of La Draga 13 Published online by Cambridge University Press
Table 1 (Continued )
Lab ID Sample Context Sector Method
ETH-88868 Cereal Structure E65 A AMS 6141 26 23.6 Event 16. Second
Beta-579521 Cereal Structure E26 A AMS 6140 30 23.3 Event 17. Second
Beta-580972 Cereal Structure E52 A AMS 6130 30 23.0 Event 18. Second
UBAR-311 Charcoal Structure E40 A CON 5970 110 Event 19. Second
ETH-88876 Cereal Structure E254 A AMS 6142 26 24.9 Event 20. Second
Beta-422871 Bos taurus Structure E260 A AMS 6210 30 18.4 4.8 Event 21. Second
Beta-428247 Sus domesticus Structure E258 A AMS 6130 30 20.8 4.5 Event 22. Spatial
Beta-422872 Cervus elaphus Structure E261 A AMS 6120 30 21.0 7.1 Event 23. Spatial
Beta-481573 Bos taurus Structure E263 A AMS 5980 30 19.94 Event 24. Last
Beta-422869 Fauna Structure E258 A AMS 6060 30 20.9 4.5 Event 25. Last
Beta-425198 Sus domesticus Structure E261 A AMS 5990 30 20.5 4.6 Event 26. Last
14 V Andreaki et al. Published online by Cambridge University Press
Table 1 (Continued )
Lab ID Sample Context Sector Method
Beta-579522 Cereal Structure E255 A AMS 5990 30 24.1 Event 27. Last
Beta-315050 Cereal Layer IV D AMS 6210 40 23.4 Event 29. Second
Beta-315051 Cereal Layer IIa D AMS 6230 40 23.7 Event 30. Second
Beta-298438 Fauna Layer III D AMS 6070 40 21.1 Event 31. Last
Beta-505896 Organic matter Peaty layer 5b C AMS 5360 30 29.7 Post-Occupation I
Beta-505895 Organic matter Peaty layer 3b C AMS 5060 30 26.2 Post-Occupation II
Beta-291443 Triticum Structure E240 D AMS 4860 40 24.1 Post-Occupation III
Absolute Chronology at the Waterlogged Site of La Draga 15 Published online by Cambridge University Press
Most samples have been retained for analysis, even with relatively large lab errors. Only when
statistical analysis proves that there is an error in the estimate, the date has been processed as an
outlier. AMS and conventional methods have been both considered, and only when
discrepancies are very clear, a separate analysis has been carried out.
Dendrochronological Analysis
The excavations carried out to date at the site, have made it possible to recover 1271 piles and
494 horizontal timbers, counting 1765 structural timber logs that can potentially be
dendrochronologically measured. The dendrochronological analysis of the piles is still in
progress (Piqu´
eetal.2021;L´opez-Bult´o et al. forthcoming). The vast majority (>95%) of
timber logs have been determined as oak wood (Quercus sp. deciduous) (Bosch et al. 2006;
opez-Bult´o and Piqu´
e2018). Other taxa identified are hazel, laurel, and dogwood.
Another main characteristic of the archaeological wooden timbers from La Draga is their
relatively small average diameter, especially the low average number of tree rings per
sample. The number of samples with less than 30 tree rings is higher than 70% (L´opez-
Bult´o and Piqu´
e2018). Many samples with higher number of tree rings have very narrow
rings, with almost no latewood, which make their measurement difficult.
So far, tree rings from 136 piles and horizontal timber logs have been described and measured,
providing a floating dendrochronological sequence that covers an interval of 265 years. The
dendrochronological sequence could not be correlated with any other, due to the absence of
an absolutely dated dendrochronological sequence covering the Neolithic period up to the
present of the northeastern part of the Iberian Peninsula.
Prehistoric inhabitants used freshtree trunks for building. The preservation of the last
growth ring (cambium. cf. Rathgeber et al. 2016) in 66% of the dendrochronologically
measured piles allows establishing a single depositional event of tree felling during the
winter of the year 237/238 of the local tree ring sequence. Eight percent of the measured
logs appear to be older, however, having been cut between winter 233/234 and winter 236/
237, and they may come from reuse, stored wood, or dead standing trees. The latter (year
236) may have the same origins as the older ones or, more plausibly, have been part of a
preparatory felling for the main site that would begin the following year.
After the year 237/238, new piles were added as reinforcement and repair of the structures
(platforms and/or dwellings). These trees were cut between three and 28 years after the first
tree-felling. Given that no other pile has a more recent tree ring, we may assume that
maintenance and repair of built structures stopped after 28 years. However, it is important
to consider that tree trunks used for repair appear to be younger and thinner than those
used for initial construction, and they are more difficult to recognize as construction
elements. There are still many thin trunks waiting for tree ring count, and therefore the
available last tree ring (265 in the local sequence) is not necessarily the last one, nor does it
represent the final event of the first occupation (Figure 6).
From the concentration of trunks cut at the same year, it appears that most wooden structures
in the prehistoric settlement were built in one year (perhaps two), during the winter of the
dendrochronological year 237/238 and during the previous year 236 of the local tree-ring
sequence. In all four excavated sectors (A, B, C, D) we found logs coming from the same
foundational forestry cut of 237/238, suggesting the strict contemporaneity of wooden
structures all along the Neolithic settlement (Figure 7). This would imply that the first
16 V Andreaki et al. Published online by Cambridge University Press
1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250 260
Construction phase of 237/238
Forestry cuttings of 237 used as a basis for the village (live wood + stored and recycled wood)
Forest stands (oak groves) harvested from 233 to 237
Oak Grove
Oak Grove
Oak Grove
Figure 6 The local dendrochronological sequence at La Draga
site. The wooden posts corresponding to both construction and
repair phases are represented in the diagrams.
Absolute Chronology at the Waterlogged Site of La Draga 17 Published online by Cambridge University Press
village of la Draga was built in one go, building wooden constructions all around the
settlement area.
The strict contemporaneity of most of the wooden elements used for construction is quite
unusual compared with other apparently contemporaneous lakeside settlements. In
Switzerland (and throughout the Alpine Arch) in sites dating to the Early Neolithic period,
settlements are generally smaller, and it is unlikely that the construction of built structures
occurred simultaneously. On the contrary, settlements grew gradually, expanding to
neighbouring areas, such as the case of Hornstaad-Hörnle IA, built from 3910 BC onwards
(Billamboz 2006), and Sutz-Lattrigen/Riedstation, built between 3393 and 3389 BC (Hafner
1994). Similar processes are also documented for more recent settlements like Cortaillod-
Est, dated in the Final Bronze period (Gassmann 1984; Arnold 1986).
Cross Dating: Dendrochronology and Radiocarbon Dating
A sample of 13 architectonical wooden elements have been dated by radiocarbon. For nine of
them, the last growth ring (cambium) has been dated by AMS. The other three dates come from
conventional radiocarbon dates of groupings of outermost rings of the same log. Seven samples
correspond to the foundational tree felling, and other six to timber logs used for the repair of
the wooden structures after that date (see Table 2).
All samples are water-saturated wood sherds, which could have altered the original
radiocarbon content. Waterlogged samples have poor cellulose preservation; although ABA
treatment should have removed possible contaminants, it is not as effective dealing with
problems that may have been caused by potential rootlets of plants living around or on the
Figure 7 Map of spatial distribution of the dated wooden piles at La Draga, including both the installation (year
237/238) and the repair associated piles. Sector A on the left and Sector BD on the right.
18 V Andreaki et al. Published online by Cambridge University Press
Table 2 Results of the dendrochronological analysis of the wooden piles and horizontal elements used either for construction or/and repair
purposes in La Draga.
Number Sectors Species
(internal chronology)
first ring date /
date of the last ring
Oak forest
Dating 14C
Method Lab ID BP sd 1 σ2σ
PT-1311 A QU 22 Winter 216/237 II-A A Beta481571 6270 30 5299 5225 5316 5211
PT-0986 D QU 36 Winter 202/237 II-A A UA62942 6285 39 5305 5225 5370 5200
PT-0089 A QU 56 Spring 182/237 II-B A UA62940 6401 38 5470 5320 5470 5310
PT-0738 B QU 85 Winter 153/237 II-C A UA62941 6308 39 5320 5225 5370 5210
PT-0605 B QU 57 Winter 181/237 II-D C UBAR1308 6270 45 5304 5219 5341 5072
TT-0468 D QU 168 Winter 38/237 II-D A UA62943 5971 41 4910 4790 4960 4720
TT-0468 D QU 168 Winter 38/237 II-D A Ua-65467 5979 37 4931 4798 4987 4732
PT-0106 A QU 59 Spring 178/237 II-A C UBAR314 6410 70 5472 5322 5481 5217
PT-1450 A QU 19 Winter 223/241 III-A A Beta505910 6210 30 5282 5066 5301 5049
PT-1441 A QU 20 Winter 228/247 III-A A Beta481572 6320 30 5357 5226 5472 5081
PT-0153 A QU 26 Winter 223/248 III-A A Beta425196 6310 30 5326 5228 5361 5223
PT-0191 A QU 28 Winter 229/256 III-B A Beta425195 6260 30 5302 5228 5315 5215
PT-0584 B QU 43 Spring 217/259 III-B C UBAR1248 6240 35 5303 5085 5308 5071
PT-0582 B QU 28 Winter 238/265 III-B C UBAR1247 6295 45 5313 5226 5374 5080
Absolute Chronology at the Waterlogged Site of La Draga 19 Published online by Cambridge University Press
wood (Haesaert et al. 2013). Since several different laboratories have been implied in dating
wood samples from La Draga, differences in pretreatment may be among the causes for
estimated ages differences, as has been suggested by other authors dealing with the same
problem (Sjögren 2011). Furthermore, there are relevant differences in the post-depositional
alteration generated by the differential accumulation of water under the wooden platforms.
Water should have been accumulated during humid winters and evaporated during dry
summers. This summer evaporation, however, would have been different in open areas
than in built areas, where the ground retained moisture (eastern part of the excavated area
of this sector). As a result, organic material from open areas between buildings, and
materials below wooden platforms would have reacted differently to waterlogged
conditions, variations in phreatic level and the differential accumulation of decomposed
organic material.
Unlike the rest of the logs, used as vertical piles, sample TT0468, the thickest preserved log, is
the only dated horizontal board, and therefore its post-depositional reaction to waterlogged
conditions would have been different.
Dendrochronological and Radiocarbon Data: Wiggle-Matching
Most radiocarbon dated wood samples associated to the same dendrochronological year 237/
238 (Beta-481571, UA-62942, UA-62940, UA-62941, UBAR-314, UBAR-1308) pass the Ward
and Wilson (1978) test. Samples Beta-505910, UA-62943,UA-65467 appear to be clear outliers.
Using the IntCal20 calibration curve, the statistical combination of the 6 radiocarbon dates
that passed the test give an estimate of 6311 ±17 BP, and a calibrated confidence interval
between 5313 cal BC and 5222 cal BC (68.3% interval). The median is situated at 5254 cal BC.
A preliminary wiggle-matching Bayesian model using most radiocarbon dates and the
dendrochronological gaps between foundational piles and the ones assigned to later repairs
has been estimated. The model has very poor agreement (A
=28.7) given the presence
of three additional outliers (UA-62940,UBAR-314, and Beta-505910). After deleting those
outliers, general agreement increases significantly (A
=136.6) (Figure 8). In practice,
Figure 8 Wiggle-matching of dendrochronological ordered piles after deleting outliers. Calculated using OxCal
4.4. and IntCal20 calibration curve.
20 V Andreaki et al. Published online by Cambridge University Press
the size of the index of agreement in a wiggle-matching model varies depending on the way
original dates over-quote or under-quote their respective lab errors, but also on how
constrained the wiggle-match sequence is by the shape of the calibration curve (Wacker
et al. 2020). At this point, it is important to remark the clear bimodality of IntCal13 and
IntCal20 calibration curves after 7100 BP, probably caused by variability of atmospheric
14C content at this time interval or by poor original sampling (Oms et al. 2016; Manen
et al. 2019; Reimer et al. 2020; Bayliss et al. 2020). This adds uncertainty to the estimation
of the calendar age of outer tree rings. For the time being, only outer tree rings have been
dated using radiocarbon. We are aware that using inner rings of some of the piles we will
have the chance to get a steeper part of the calibration curve into our wiggle-match.
Dendrochronological analysis is not yet finished and when more samples get extracted, the
more they will spread outand the better the resulting precision will be. This can be
achieved by minimizing the number of possible positions where the distribution of
radiocarbon dates can match the calibration curve. For the moment, existing dates only
serve as an initial hypothesis.
Our best model suggests an estimated date around 5293 cal BC (median of the 5312-5233 cal
BC at 68.3%) for the precise moment original trees were logged (year 237/238). Tree ring 247
cannot be differentiated from tree ring 237/238. However, the model seems to differentiate
successive tree rings correctly: TR248: 5284 cal BC, TR256: 5283 cal BC, TR259: 5275 cal
BC. The last well documented repair (Tree ring at the year 265) has been documented at
an estimated date around 5272 cal BC (median of the 52915212 cal BC at 68.3%).
Dating the First Occupation at La Draga
Wiggle-matching only allows for an estimate of the temporality of depositional events 1, 2, and
3, based on isotopic events associated with the last growth ring of piles used for construction
and posterior repair. There are additional radiocarbon dated samples that have been
stratigraphically associated with this first occupation (from Depositional Events 4, 5 and
6a, 7a, 8a). They do not pass the Ward and Wilson test and therefore we cannot assume
directly the strict contemporaneity of all archaeological deposited material from this first
occupation. Nonetheless, its duration was presumibly very short given the small number of
tree rings between piles used for the initial foundation and the youngest piles used for
repair (approximately 30 years).
In Sector A, Depositional Event 4 is statistically coherent around 5266 cal BC (median of the
53035217 confidence interval 68.3%), well within the most probable period of use of the
wooden platforms estimated by the wiggle-matched model.
In Sector B, all dated samples from Depositional Event 5 pass the Ward and Wilson test and
appear to be strictly contemporaneous with sampled dates from Depositional Event 8a. A
combined date around 5271 cal BC (52985216 cal BC, 68.3% confidence interval) seems
to be a good estimate for the moment of platform use and successive repairs. Although
sample Beta-315052 from Depositional Event 6a seems to be older, all dates related with
platform use and repairs from Sectors B and D pass the Ward and Wilson test. The same
can be said about contemporaneous dated samples from Sector A.
Two dated samples come from Depositional event 7a from underwater Sector C and pass the
Ward and Wilson test. One of these, however, Beta-27856, becomes an outlier when compared
Absolute Chronology at the Waterlogged Site of La Draga 21 Published online by Cambridge University Press
with all contemporary sampled dates related with this moment of use and repair of wooden
Dated samples from depositional events 6b, 8b from Sector B and D, and depositional event 7b
from Sector C, show recurrent estimates much younger than any sampled date from this first
occupation, related with the use and repair of wooden platforms. All of them pass the Ward
and Wilson test and give a combined estimate 68.3% confidence interval around 52065032 cal
BC. They cannot be used to suggest the moment of use or repair of the wooden platform,
although stratigraphically they are in close connection with horizontal timber boards. This
apparent contradiction could be solved by considering hypothetically two differentiated
moments of platform use, one centered around 5270 cal BC and the other around 5097 cal
BC (median of the 68.3% confidence interval 52065030 cal BC). However, this hypothesis
contradicts with dendrochronological data and the results of high-resolution spatial analysis
(N. Morera, forthcoming) suggesting a single and relatively short occupation of no more
than 30/40 years.
The reasons for this apparent chronological difference within the first occupation lie in post-
depositional factors (Andreaki 2022). A pre- and post-depositional subsidence of the original
surface (Iriarte et al. 2014) observed in Sectors D and B but not at Sector A- altered the original
deposition at those areas. To test partially this post-depositional alteration hypothesis, we have
compared differences in radiocarbon calibrated dates depending on the material of the dated
sample. Wooden objects (Beta-137197 and Beta-137198) are clearly older than seeds coming
from the same stratigraphic layer NAVII (Figure 9). Faunal samples show a clear stratigraphic
inversion, where samples found at the top of the archaeological layer NAVII (Beta-588213;
Depositional Event 8a), appear to be older than samples found at the bottom, in contact
with the original lake substrate (Beta-588214; Depositional Event 8b). Social life occurred
on platforms, but also on the ground surface, therefore material elements may have been
deposited above horizontally disposed boards and below them. Puddled water below
wooden platforms may have caused the slow sinking of material fallen from the platform
during use (cereal grains and charcoals) (Andreaki 2022), but there is also the possibility of
the intentional anthropic burial of materials. No chronological difference should exist
between samples found in contact with the preserved wooden elements, and those under
them, found in contact with the original lake marl surface. The spatial microanalysis of
animal bones gives partial support to this hypothesis (Morera et al. 2019).
It is then very difficult to identify the very last moment of this first occupation associated to the
wooden structures. Some elements coming from the second occupation may have been
infiltrated into the older occupation debris because of indirect pressure from the
accumulation of travertine slabs above. Some light charcoal and seed samples would have
floated as they felt in the water table, and as a result would not get sunk immediately, as in
the case of heavier wooden artefacts. Furthermore, Depositional Event 28 appears to be a
small pit excavated penecontemporaneously or after the start of the second occupation, and
stratigraphically affecting the ground below.
The end of La Dragas First Occupation and the probable abandonment of the site for some
years can be explained by a combination of environmental changes that occurred on the shores
of Lake Banyoles, given intense forest exploitation and the geomorphological evolution of the
lake shoreline. Although some short and temporal reoccupations cannot be excluded given the
22 V Andreaki et al. Published online by Cambridge University Press
number of piles that have not yet been dendrochronologically dated, a subsidence of the ground
surface of those parts of the settlement nearer to the shoreline marks a probable interruption of
social life at the site, as the micromophological data suggest. Those analysis suggest that the
ground subsidence was already present before the first occupation, to a lower degree, although
it would probably not have been perceptible to the inhabitants of La Draga. In fact, the earlier
beginning of this subsidence at a smaller degree is what made possible the accumulation of
organic materials in Sectors BD and the formation-preservation of peaty strata. Further
subsidence during the following years and the parallel increase of the water table in this
sector of the site may have been the reason of its abandonment. The constant water
presence in this area would have also delayed the sites re-occupation for some time.
Figure 9 Comparison of wood, seeds and faunal samples from Sectors B and D, First Occupation. Using Oxcal 4.4
and IntCal20 calibration curve. In all three cases, R_Combine has been used to create the posterior distribution. In
the case of wood and seed samples, this procedure pass Ward and Wilson test. In the case of faunal samples, there is
an outlier (Beta-278255), 100 radiocarbon years older than the rest. See discussion in text.
Absolute Chronology at the Waterlogged Site of La Draga 23 Published online by Cambridge University Press
Dating the Second Occupation at La Draga
There is a possible gap of approximately 100 years (or a bit more) between the collapse of the
wooden structures and the beginning of a new occupation, a time interval in which the site may
have been abandoned. Signs of exposure have been detected microscopically in the sediment of
Sector D just above the preserved timber planks, highlighting a period of exposure before the
arrangements for the new occupation (Andreaki 2022). Pollen data retrieved from the
beginning of the new occupation suggest that the forests around the new settlement would
have experimented a clear recovery after a period of local deforestation during the first
occupation (Revelles 2017,2021; Revelles et al. 2017). 100 years seems a likely estimate for
the time interval during which the forest recovered.
This new occupation is associated with the arrangement of travertine slabs forming a
pavement, perhaps as a kind of insulation platform to reduce the passage of groundwater
and isolate the occupation floor from the mud. The new occupation would have been
longer than the first, and some different moments can be distinguished (Andreaki 2022).
In Sector A, the first and second occupations are clearly differentiated. A two sequential phases
OxCal model gives a A
=134.9 after deleting three outliers. The end of the first occupation
can be estimated around 5252 cal BC (median of the 68.3% confidence interval), a transition
between both phases around 52115199 cal BC, and the beginning of the new occupation after
a probable hiatus of more than 50 years, well attested in the settlement areas less affected by the
changing levels of the lakes water table.
Material from hearths in Sector A (Depositional Events 9-19), made of travertine slabs located
in close stratigraphic contact with the original lake marl surface, is the oldest from this
occupation. All available dated samples (bones and seeds) have been combined after
passing successfully the Ward and Wilson test, suggesting a date around 5066 cal BC
(median of the 52055013 cal BC, 68.3% confidence interval). Two cereal samples from
Sector D (Beta-315050,Beta-315051) are statistically contemporaneous. They come from
two different Depositional Events: 29 and 30. Depositional Event 29 is a clay layer with
high presence of organic material immediately above wooden collapsed materials from
previous occupation. Depositional Event 30 is a sediment of dark color, with little presence
of archaeological material and mostly consisted of decayed organic matter, whose inferior
part is in contact with the travertine paved area (Andreaki 2022). If we consider only the
oldest dates from those depositional events (Beta-422871,Beta-315050 and Beta-315051), a
date around 5140 cal BC would be a good preliminary estimate for the beginning of the
new occupation. The reoccupation of Sectors D (near the lake shoreline) and Sector A (700
m away) would have been contemporaneous.
In Sector A, a new and later rearrangement of settlement areas can be suggested given the
statistical difference between dates from the bottom of structures E258 and E261 (Beta-
428247, Beta-422872) (Depositional Events 22, 23) and dates sampled at the top of the
sediment filling those structures (Beta-422869, Beta-425198) (Depositional Events 25, 26).
The oldest dates pass the Ward and Wilson test and can be considered strictly
contemporaneous within the interval 50435007 cal BC. Younger dates from the same
structures, together with other dates from samples found at E263, and the top part of
structures E258 and E261 are clearly later (Depositional Events 24, 27), and they would be
dated around 49304882 cal BC (68.3% confidence interval), suggesting a possible
modification of previous structures. Event 31, from Sector D, a sample of fauna from a
24 V Andreaki et al. Published online by Cambridge University Press
sandy greyish sediment containing travertine sand and fragmented slabs, and stratigraphically
correlated with the upper part of the travertine slabs found in the paved area, would also belong
to this very last occupation.
Radiocarbon dates from underwater Sector C do not change the broader conclusions. Four
faunal bones and cereal seeds come from this sector. Statistically, the four radiocarbon
dates do not pass the Ward and Wilson (1978) test. The stratigraphically deepest dated
sample appears to be older (68.3% confidence interval: 53725067 cal BC) than samples
coming from superior layers (68% confidence interval: 52164981 cal BC). Dates show
however a relevant degree of chronological overlapping.
Global Chronological Model
A detailed chronological model has been calculated based on the assumption of 9 differentiated
phases(Figure 10). The first one integrates the original tree-felling and the pile-dwellings
construction (Phase Construction, Depositional Events 1, 2, and 3), whereas the Phase
Use and Repairbrings together all the samples associated to wooden structures use and
repair (Depositional Events 4, 5, 6a, 7a, 8a). A single outlier (Beta-425194) has been
deleted. These first two phases have TPQ and a TAQ constraints based on the results of
the previous wiggle-matched model based on the difference in the number of tree rings
between logs used for construction and logs used in later repairs from 5290 cal BC until
5250 cal BC. Both estimates have 10 years of standard error.
After a gap of 50 years representing a moment of abandonment of the local area, a
Transitional Phasecollects samples mostly from first occupation that experimented some
form of stratigraphic inversion as a consequence of karstic subsidence and changing levels
in phreatic level (Depositional Events 6b, 7b, 8b, 28). During this chronologically separated
transition phase, reuse of space is not excluded.
Figure 10 Results of a model of 9 partially contiguous, partially sequential and partially overlapping phases and
sequences. Oxcal 4.4. IntCal20 calibration curve.
Absolute Chronology at the Waterlogged Site of La Draga 25 Published online by Cambridge University Press
The Second Occupation is analyzed as configured by three different moments: Second
Occupation(Depositional Events 921 from Sector A, Depositional Events 29 and 30
from Sector D), Spatial Rearrangement(Depositional Events 22 and 23 from Sector A),
and Last Neolithic Occupation(Depositional Events 2427 from Sector A, and
Depositional Event 31 from Sector D). Three posterior phases distinguish the dated
samples found in more recent layers, affected by ancient and modern erosion.
A preliminary OxCal implementation of this model distinguish four outliers: HD-15451,
UBAR-313,UBAR-311, too modern, although global model agreement is very high
=108 A
=101). UBAR-311 is a problematic non-AMS sample, with an
excessively long standard lab error. UBAR-313 is another non-AMS dated sample, and
HD-15451 is an isolated finding. Beta-315051 is another problematic date. Initially
considered characteristic of the second occupation, it was found very deep in the
stratigraphic sequence (NAIV), but in close contact with travertine samples above. It seems
much more related with use and repair of wooden structures than with later phases. We
have deleted those dates and executed the model again, with a huge gain in agreement
=167.8, A
According to this model (Figure 10, Table 3), a first occupation on the shores of lake Banyoles
can be placed along the temporal interval of 53025247 cal BC. The hiatus in which areas of the
settlement were probably abandoned, although temporal and short reoccupations cannot be
excluded would have arrived until ca. 5100 cal BC. The depositional events integrated into
the so-called Transitional Phase appear to be later that the most probable start of the
Second Occupation. On one hand, the oldest dates for second occupation (Beta-422871 and
Beta-315050), around 5200 cal BC, are a priori too old for dating properly the moment the
original ground was insulated with travertine slabs. The remaining dates are grouped
around a median of 5075 cal BC. The best hypothesis would be to make emphasis on a
relatively long period of abandonment (100 years) and a relatively later reoccupation with
restructuring of the ground surface. On the other hand, most dated samples from the
Transitional Phase are small seeds between the remains of the wooden debris and the
travertine slabs, that may have been affected from the second occupation and the influence
of groundwater.
The second occupation would be longer than the first one, from 5075 cal BC until 4860 cal BC.
Two successive settlement rearrangements may be suggested, the first around 50615025 cal
BC and the second and last one around 49174821 cal BC. Very few remains of occupation
exist for the period after 4800 cal BC.
ChronoModel 2.0 ( has a different way to calculate the a posteriori
temporal intervals. It is based on the concept of Event (Figure 11). An Event is a point in
time for which a hierarchical Bayesian statistical model can be defined (Lanos and
Dufresne 2019). In our case, it corresponds to what we have defined as a depositional
event, when the temporal duration can be argued as less than the lab error of the isotopic
date. A Phase is a group of Events, and it is here defined as a series of related depositional
events, whose joint temporal duration exceeds 3040 years. The modeling approach is very
different between OxCal and ChronoModel, although in both cases the temporal duration
of phases is estimated in terms of the difference between a start and an end event. We have
used exactly the same number and definition of phases in the models implemented in
OxCal and ChronoModel. ChronoModel differs from the OxCal model in the way
26 V Andreaki et al. Published online by Cambridge University Press
depositional events and their stratigraphic anterior/posterior constraints have been included.
Unlike Event model, the Phase does not respond to a statistical model: indeed, we do not
know how events can be a priori distributed in a phase. However, we may question the
beginning, end or duration of a phase from the Events that are observed there (query). A
level of a priori information can be added: the Events from one phase may be constrained
by a known duration and a hiatus between two phases can be inserted (this imposes a
temporal order between two groups of Events). In ChronoModel, constraints link events
and not calibrated dates. (Lanos et al. 2016; Lanos and Philippe 2017).
The idea is to estimate the unknown date of phases based on dated samples associated to
Events, which in their turn, are associated to Phases. The event model, implemented
in ChronoModel, combines contemporary dates, t
::: t
, with individual errors, σ
::: σ
in order to estimate the unknown calendar date θ. The following equation shows the
stochastic relationship between t
and θ:
where εCM
i~N0;1for i=1to nand εCM
nare independent. θis the unknown parameter
of interest and σ
::: σ
are the unknown standard deviation parameters. That means that each
parameter t
can be affected by errors σ
coming from different sources (Lanos and
Philippe 2017).
The temporal position of each phase on the calendar scale is estimated according to the events
included in it. The following information are given for each phase:
The beginning of a phase, α, reflects the minimum of the revents included in the phase:
Table 3 Results of La Dragas Chronological 9 Phases Model (OxCal 4.4.) after outlier
elimination (68.3% confidence interval).
Phase From To Median
First occupation
Tree ring 237/238 C_Date (5290,10) 5309 5290 5299
Platform construction starts 5302 5277 5289
Platform use and repair ends 5272 5247 5260
Tree ring 265 C_Date (5250,10) 5258 5239 5248
Transitional phase
Transition starts 5095 5051 5076
Transition ends 5031 4975 4997
Second occupation
Second occupation starts 5207 5048 5075
Settlement rearrangement starts 5061 5025 5045
Last Neolithic occupation ends 4917 4821 4862
Post-occupation I starts 4779 4176
Post-occupation II starts 4479 3824
Post-occupation III starts 4564 3553
Post-occupation III ends 3684 2759
Absolute Chronology at the Waterlogged Site of La Draga 27 Published online by Cambridge University Press
The end of a phase, β, reflects the maximum of the revents included in the phase:
The duration, τ, is the time between the beginning and the end of a phase:
The posterior distribution of all these elements may be approximated by MCMC methods and
statistical results such as the median, the standard deviation and so on, may be estimated
(Lanos and Dufresne 2019).
The idea is to reduce the uncertainty of calendar date estimates using hypothesis of starts and
ends based on stratigraphic relationships, or the number of tree rings between different
radiocarbon samples of the same tree-ring sequence. Both computer programs apply
Bayesian probability reasoning to define the proper limits of their assumed broad
contemporaneity in terms of the spread of the dates, and the interphases or boundary
temporal limits of the phase. In ChronoModel, the radiometric date of an event is assumed
to be affected by an unknown error sigma, which will be estimated a posteriori. If this error
is too large, compared to the error of other dates, we will be dealing with an outlier,and
in this case, OxCal would have displayed an A
lower than 60%. On the other side, in
ChronoModel, there is no need to remove this date: it will be automatically discounted
because of this high individual posterior error (Lanos and Philippe 2018,2020).
Consequently, in ChronoModel, there is no sorting of dates according to outlier
elimination steps: all dates are considered, but some of them are later discounted, and as a
result, do not affect the phase temporal range when they diverge from the other dates
(Lanos, personal communication).
To reproduce exactly the model previously estimated using OxCal, two temporal bounds were
included in our ChronoModel estimation (Figure 11), representing the dendrochronologically
deduced temporal limits: 5293 cal BC for the beginning of the process (tree felling), and 5272
cal BC for the last documented repair. We have added 10 years in this last case concerning the
uncertainty of the last moments of wooden platforms occupation. After that bound we have
defined a Transition Phase,with an additional uncertainty of 100 years, and a gap of 50 years
before the beginning of the second occupation. This is exactly the same model we defined using
OxCal 4.4. The second occupation is differentiated from the later spatial arrangement and the
last evidence of a Neolithic Occupation. Post-Occupation evidence has been integrated into
three different phases.
MCMC has been configured with 1000 burn-in iterations, 500 further iterations for the
adapting cycle, and finally 100,000 iterations for the final acquisition of posterior
distributions (thinning =10). Gelman-Rubin test is not yet implemented in the current
version of ChronoModel 2.0. Consequently, we have checked the MCMC convergence
visually by locking at the stability of autocorrelation plots. We have also checked the
acceptance rate at 44% in the case of a Metropolis-Hastings with a Gaussian random walk
(Roberts et al. 2001), and the decorrelation of the variables (Lanos, personal communication).
The same outliers that were deleted in the OxCal model were eliminated, once we have checked
that acceptance rate in ChronoModel for those dates was also around the 44% threshold.
28 V Andreaki et al. Published online by Cambridge University Press
Figure 11 An Event and Phase Model with stratigraphic constraints and temporal boundaries based
on dendrochonological and Wiggle-Matching estimates. ChronoModel 2.0. IntCal20 calibration curve.
Absolute Chronology at the Waterlogged Site of La Draga 29 Published online by Cambridge University Press
Results obtained by ChronoModel are in good consonance with those obtained with OxCal
(Figure 12,Table4). Both models give support to the hypothesis of a hiatus of nearly 100
years between the end of the first occupation and the beginning of the second, but precise
estimates for their start and end slightly differ. It is important to consider, however, that this
is not an occupation phase, with clear-cut start and end, but a region of temporal uncertainty
where post-depositional processes affected previous and posterior stratigraphic layers.
Both models agree with an estimate for the beginning of the second occupation around 5090 cal
BC. Nearly 70 years after, there is evidence of a spatial arrangement of built spaces. According
to both models, the last evidence of Neolithic occupation was not later than 4780 cal BC.
Figure 12 Results of a 9 Phases Chronological model. ChronoModel 2.0. IntCal20. Each Phase is depicted with a
different color. The lighter color corresponds to the a priori confidence interval, whereas the darker color depicts the
a posteriori Bayesian estimation.
30 V Andreaki et al. Published online by Cambridge University Press
Using the above stratigraphical ordering and radiocarbon estimates for the duration of
depositional events, a general temporal sequence was defined based on Allen algebra
estimated temporal relationships (Allen 1983; Zoghlami et al. 2012; Dye and Buck 2015;
Belussi and Migliorini 2017; Drap et al. 2017; Barcel ´o and Andreaki 2020) (Figure 13).
The period of transition between the first and second occupation remains in the temporal
interval from 5200 to 5100 cal BC, although there is a clear overlapping with depositional
events from the second occupation (Depositional Events 29 and 21).
La Draga in the Chronological Context of the Early Neolithic of the Western Mediterranean
Very few traces of pre-farming, Hunter-Gatherer Mesolithic occupations, are known for the
Lake Banyoles area. At Bauma del Serrat del Pont (Tortellà) (Alcalde et al. 2009; Alcalde
and Sa˜na 2017), 20 km from La Draga site, in a slightly different ecological niche, a
human occupation dated around 73806000 cal BC has been identified. Its material culture
can be attributed to a local technological tradition referred to as notches and denticulate
Mesolithic(Martínez-Grau et al. 2020), which developed locally in NE Iberian Peninsula
and extended from the Pyrenees to the Mediterranean seaboard and in the Ebro basin.
Bauma del Serrat del Pont is a small rock shelter, and it was occupied repeatedly during
this period, with at least five different seasonal occupations registered, the last one
extending until ca. 6000 cal BC. The archaeological record suggests a recurrence of short
occupations by small groups, probably between the end of summer and the beginning of
winter, practicing a very homogenous strategy based on intensive exploitation of medium
and large mammals. There is evidence derived by marine valves concerning river fishing
and travels to the Mediterranean coast (Alcalde and Sa ˜na 2017).
Table 4 Results of La Dragas Chronological 9 Phases Model (Chronomodel) after the
elimination of outliers. (HPD and Phase Time Range 68.3%).
Phase From To Median
First occupation
Tree Ring 237/238 C_Date (5290,10) 5293
Events 1, 2 and 3 start 5293 5290 5291
Events 4, 5, 6a, 7a, 8a end 5275 5272 5274
Tree Ring 265 C_Date (5250,10) 5272
Transitional phase
Events 6b, 7b, 8b, 28 start 5209 5149 5174
Events 6b, 7b, 8b, 28 end 5135 5069 5104
Second occupation
Events 921, events 29 start 5111 5052 5084
Events 22, 23 start 5042 4984 5012
Events 24, 25, 26, 27, 31 end 4868 4729 4786
Post-occupation I 4324 4084
Post-occupation II 3945 3794
Post-occupation III 3705 3545
Absolute Chronology at the Waterlogged Site of La Draga 31 Published online by Cambridge University Press
The abandonment of Mesolithic occupations in the area near the Banyoles lake is not an
isolated historical event. Studies based on the sum of probability intervals of radiocarbon
dates have suggested a decay in the number of known sites between 5900 and 5700 cal BC
(Balsera et al. 2015; Bernabeu et al. 2016; Bernabeu et al. 2017; Fyfe et al. 2019; Pardo-
Gord´o and Barcel ´o 2020; Martínez-Grau et al. 2021). Bernabeu et al. (2017) have linked
this occupational interruption with dry and cold intervals having occurred regularly on the
area. One of them would have happened before the arrival of farmers, possibly related with
the North Atlantic ice rafting debris 5b IRD event, dated around 5900 cal BP (Frigola
et al. 2007; Wanner et al. 2011; Finn´
eetal.2019). We must be careful about using these
climatic anomalies to explain changes in the archaeological record. The Banyoles lake is in
a humid region and such an event would not have implied dramatic changes in the
landscape, as might have happened in more arid regions such as the southeast of the
Iberian Peninsula. We must bear in mind that in the case of the lake, a phase of higher
humidity or an increase in rainfall (lake flooding and inundation) would probably have
more effect than aridity. With available data, we cannot be certain whether a temporal
increment in aridity caused variations on the lake shoreline and the phreatic water level to
explain the probable abandonment of the settlement for a while. The last volcanic
eruptions in the neighboring volcanic area of Garrotxa-Olot20 kms from the lakemay
have also some impact on depopulation (Maria Sa˜na, personal communication).
In stratigraphic continuity with the last Mesolithic occupation, but nearly 500 years after
(54705380 cal BC), a new occupation at Bauma del Serrat del Pont differentiates from
Figure 13 Allen Algebra adapted diagram for site occupational temporal sequence representing
depositional events E-1 to E-31 and their respective phases. Black vertical lines mark the end of
clear evidence from first occupation (5200 cal BC), as well as the beginning (5100 cal BC) and end
of second occupation (4800 cal BC).
32 V Andreaki et al. Published online by Cambridge University Press