1.1. Material and methods

1.2. Results

8For Communal building St 10, all the samples came from the substructures (embankment or stony floor) or from the built floors or benches of the three earlier cycles or the construction/demolition (chap. 5). The dates (N = 27) were very consistent and range from 9690 ± 50 to 9380 ± 45 BP. Only one sample from SU 10.6 gave a significantly younger age (8490 ± 40 BP). This result was not confirmed by the other four samples from this SU with ages ranging from 9590 ± 40 to 9450 ± 60 BP. This sample is therefore considered an outlier and was not included in our analysis.

9The dates of Sector B (N = 21) all came from the foundation trenches of well-identified buildings. They were all consistent, with the exception of a modern outlier (215 ± 30 BP, in B15, a highly degraded and superficial building; chap. 7), and vary between 9425 ± 35 and 9590 ± 35 BP.

10The two St 800 samples prepared by the Belfast lab gave ages significantly (300 to 400 years) older than the rest of the sequence: 9833 ± 54 BP and 9869 ± 46 BP. These samples were then re-dated by ARTEMIS and provided results (9525 ± 45 BP and 9470 ± 45 BP) that were consistent with the rest of the sequence. Following an email exchange with the Belfast team, the small difference in background (48,000 vs. 50,000 years) was insufficient in explaining these older ages. They stated that because of the fragility of the samples they had applied a very light acid pre-treatment, though this was probably insufficient to remove carbonates; however, in the absence of certainty, we decided not to consider these results in the general discussion.

11Altogether, the 52 validated measurements form a very coherent series spanning less than 500 ¹⁴C years (9870-9380 BP; table 10-1).

Tab. 10-1 – Radiocarbon dating of seeds and charcoal from the Klimonas site.

12Given the low dispersion of the values, and the long duration of the analysis period, it was crucial to check the reproducibility of our results over the long term; therefore, we measured two samples twice, one year apart. These samples were taken from buildings B10 and B27 in Sector B. The absence of significant difference observed on the two duplicates (St 6602: 9565 ± 45 BP vs. 9555 ± 35 BP; St 6830: 9580 ± 45 BP vs. 9580 ± 35 BP) allowed us to compare the measurements carried out over the duration of the whole project.

13The calibration of the ¹⁴C ages was obtained using Oxcal. 4.2 software in conjunction with the IntCal 13 calibration curve (Reimer et al. 2013). For each sample, the calibrated interval was relatively wide (300-400 years) due to the presence of a calibration plateau centred around 9550-9600 BP (fig. 10-1; Manning et al. 2010). If we take the upper and lower bounds of the dataset, we obtained a wide time interval of nearly half a millennium, between 9180 and 8624 cal BC (2 σ). It is therefore clear that the calibration plateau artificially increases the duration of occupation of the site. In order to counter this bias and to propose a more realistic estimate of the duration of occupation of the site, we modelled the charcoal dataset using a Bayesian approach.

Fig. 10-1 – Calibration probabilities distribution of the 52 radiocarbon dates obtained for Klimonas (light grey outlines) and their chronological position with reference to the general radiocarbon calibration curve (blue) after Reimer et al. (2020). OxCal4.4 and IntCal20.

2.1. Used dates and a priori hypothesis

14Due to the low number of dates obtained for St 800 (N = 2) and the lack of a clear stratigraphic link with the other structures, it was not possible to incorporate them into the model. The last dates made on seeds (ECHo in tab. 10-1) were also not taken into consideration in the model. The two sequences that could be modelled using the Bayesian approach were the Communal building St 10 and Sector B. In the absence of a clear stratigraphic link between these two areas, we modelled the occupation period of each of the two sectors separately.

15For the St 10 Communal building, the samples came from the three older cycles or construction/demolition (chap. 5):

1. Soil and occupation related features from “Building 1”: SU 10.10, 10.14, 10.44 and St 64 (N = 7);
2. Substructure fills (SU 10.3) and demolition levels (US 10.8) of “Building 2” (formerly named “intermediate building” in Vigne et al. 2012, 2017b; N = 6);
3. Unformal floor of “Building 2”: SU 10.6 (N = 5);
4. Formal floor (SU 10.2) and bench (St 153) of “Building 3” (formerly named “Building 2”, in Vigne et al. 2012, 2017b; N = 8).

16For Sector B, as the samples all came from well-identified building foundation trenches and as most of them were cutting one another, using these stratigraphic relationships that were discussed in chap. 7 we were able to create Harris matrices. In both the two most likely matrices (fig. 10-2, A-B), the six (if we except B15) dated buildings were ranked in the following chronological succession, from the oldest to the youngest (hypothesis H1): B27-B26-B13-B14-B29, B09 being considered as the youngest building of the terrace (chap. 7). However, due to the high degree of intertwining and erosion of the building, and to large pits which partially blurred the intercutting between B13-14-29 and B26-27, a small doubt subsisted about the position of B26. A third less likely scenario was therefore considered for the Bayesian modelling (hypothesis H2): B27- B13-B14-B29-B26, B09 (fig. 10-2, C).

Fig. 10-2 – Harris matrices of the buildings of Sector B: A and B, The two most likely chronological correlations as proposed in chap. 7 (Hypothesis H1 for the Bayesian model). B, less likely alternative (H2). Yellow boxes indicate buildings for which 14C measurements were performed. The numbers in brackets indicate the number of measurements for each building.

2.2. Method

17The dates were digitalised into Oxcal.4.3 software (Bronk Ramsey 2009). As the age differences per layer were less than the uncertainty for the ages, we sequenced these phases by postulating continuous phasing (without abandonment period) between the four different stages of the occupation of St 10 and the six Sector B phases.

2.3. Presentation and discussion of the v0 models

18As it stands, the three v0 models (model v0_St10, v0_sectorB_H1 and v0_sectorB_H2; appendix 10-1) were weakly supported (Amodel = 20, 50 and 50); therefore, the scheme proposed a priori was not validated. This may be due to the fact that some samples were not in place in the proposed stratigraphy, or were not contemporary with the events we wished to date (old wood effect). In OxCal software, a model becomes plausible as soon as its Amodel becomes greater than 60. It is interesting to note, therefore, that the models corresponding to hypotheses H1 and H2 are equiprobable, and it is not possible to favour one hypothesis over another on strictly radiometric grounds. Due to this, we decided to retain only hypothesis H1, considered more probable from an archaeological point of view (chap. 7), and to treat the two sequences corresponding to St 10 and Sector B within the same model.

2.4. Outlier-Models with a 5% probability error

2.4.3. Exponential outlier model (old wood effect)

21We then tested an outlier model based on an exponential law, to simulate a possible old wood effect. Such a model took into account a priori the possibility that all samples were a little too old, and therefore calculated the outlier probability a posteriori following an exponential growth law. There were at least two forms of pistachio tree in the region (chap. 23); one is a tree and can potentially live quite a long time, while the other is a shrub. Unfortunately, it was not possible to determine the charcoal to species level, which leaves some uncertainty about the size of this possible old-growth effect. In case of doubt, we assumed between 10 and 100 years of age for all samples and an a priori outlier probability of 100% (i.e. we assume, a priori, that all the charcoal were older than the true age of the layers) (appendix 10-3).

22The a posteriori outlier distribution indicated a high probability density for an old wood effect between 0-22 years (1 σ) or 0-91 years (2 σ; fig. 10-3).

Fig. 10-3 – A posteriori distribution of charcoal age in the v0_Exp100 model. OxCal v4.3.2 Bronk Ramsey (2017); r:5. IntCal13.

2.5. Occupation duration of the village

23Table 10-2 summarises the start and end dates for the Communal building and Sector B for the two models (General and Exponential). The second model being the most robust, we propose to date the occupation of both the Communal building and the Sector B between 8837 and 8722 modelled BC.

Tab. 10-2 – Summary of the start and end occupancy dates for the Communal building and the Sector B, according to the analytical data appendix 10-2 and 10-3).

24Whatever the model used (General or Exponential), the duration of the site’s occupation (or, more strictly speaking, of the part that we were able to date) was very short, between 0 and 60 years for the two sectors, with a strong probability (1 σ) that this duration was between 0 and 20 years (fig. 10-4).

Fig. 10-4 – A, occupation sequence of the Klimonas site modelled using Oxcal 4.3 for the v0_OutlierExp100 model; B, duration of the occupation for the v0_Exponential100 model. OxCal v4.3.2 Bronk Ramsey (2017); r:5 IntCal13 atmospheric curve (Reimer et al. 2013).

25We also tested the probability that the start and end dates of the occupation of the Communal building and Sector B were different, using the Order function of Oxcal. Whatever the model (General or Exponential) the start and end dates of occupation of the two sectors were not significantly different (p > 0.05). The most parsimonious hypothesis is therefore that these two sectors are contemporary.