Version classiqueVersion mobile

The Human Face of Radiocarbon

Zoï Tsirtsoni

Chapter 2. “Balkans 4000”: the radiocarbon dates from archaeological contexts

Yannis Maniatis, Christine Oberlin et Zoï Tsirtsoni

Texte intégral

  • 4 Supra, chapter 1, sp. p. 36‑37, fig. 2, and table 2.

1In the previous chapter we exposed the general framework in which the “Balkans 4000” project was conceived and the main criteria used for the choice of sites to be dated4. Here we discuss the precise conditions under which the selection and processing of samples were conducted, and present the results in detail.

Sampling strategy

  • 5 See McFadgen 1982; Ashmore 1999.

2The first condition for a reliable radiocarbon date concerns the physical quality of the measured sample (its nature, mass, and state of preservation), the second its archaeological value (pureness, proximity to the dated events, nature of the archaeological context). In other words, one has to know exactly what is dated, and be sure that the organic material subjected to measurement represents indeed, as closely as possible, the event to which it is theoretically attributed5. For this reason we preferred short-lived samples, i.e. animal or human bones and teeth, charred grains and fruits, or small charcoal fragments that probably come from firewood (usually small tree or bush branches), rather than samples that are potentially long-living, i.e. fragments from building elements such as big posts or beams that could be old or reused. In the case of posts, an additional risk concerns the difficulty for the archaeologist in determining their precise starting point, in other words the building level to which they are truly associated, and which can be very different from the one where the piece of wood is actually found. Thus, in general, we have avoided charcoal from posts, unless nothing else was available, or in the case of sites with only one phase of occupation.

3As a rule, we preferred to take samples from in situ layers – and whenever possible, from in situ structures – rather than from sediments that risked being disturbed or redeposited. In some cases however we consciously dated samples from colluvia, i.e. from sediments accumulated at the lower part of a slope as a result of erosion. These dates provide only a terminus post quem for the depositional event properly speaking, since the samples belong most probably to the layer (or the layers) which stood originally higher on the slope and whose disintegration formed the colluvium. The important thing, once again, was to know that these were indeed redeposited sediments, and thus avoid misinterpreting the results. We also tried to avoid samples from pits, unless their stratigraphic position was clearly determined and the character of their filling (single event, repetitive or long-term use) was well-established. Some “bad surprises” proved that one can never be cautious enough with this kind of features.

4In order to be able to select samples that met the above rigorous physical and contextual criteria, we chose to work whenever this was possible with recent excavations, and even with ongoing projects, with the direct participation of the excavators. Furthermore, this allowed the feedback of the radiocarbon dates to the excavators during fieldwork, hence providing the opportunity for modifications in the excavation strategy and/or for further and more targeted sampling. In the rare cases where samples were taken from older excavations (e.g. Kastri, Yagodina), this was done by the excavators themselves, except for a unique case (one sample from Hotnitsa), which proved however to be perfectly reliable, i.e. compatible with previous measurements from the same context.

  • 6 Supra, chapter 1, p. 32.

5In total, we selected 202 samples from 34 sites, from which we finally measured 192 (table 1). This corresponds to an increase of ca 35% of the number of dates available for the Late Neolithic/Chalcolithic and the start of the Early Bronze Age in Greece and Bulgaria6. Their internal distribution is not uniform: some sites are only represented by one or two samples, whereas others provided ten or more. Their number depended not only on the length of use of each site and the availability and quality of samples (burned-unburned layers, eroded or washed out deposits, etc.), but also on the proportion of the undated parts in the sequence. Thus, in sites with already well-established long chronological sequences, we proceeded to just a few targeted samplings, whereas others, with shorter sequences but with no prior information, called for a more abundant sampling.

Processing and measurement of the samples

  • 7 See also Evin & Oberlin 2005, p. 84‑88; Maniatis 2013, p. 6.

6According to their provenance (Greece or Bulgaria) and their initial amounts (from very small, 10‑100 milligrams, to large ones, 1‑100 grams), the samples were sent to different laboratories for suitable processing and faster production of dates by work apportionment. Three radiocarbon laboratories were involved, using three different versions of the radiocarbon measuring technique7:

    • 8 Maniatis 2013, p. 7‑10.

    The Laboratory of Archaeometry of the National Centre for Scientific Research “Demokritos” in Athens, which uses the radiometric Gas Proportional Counting technique (GPC). This method consists of converting the sample into gas (CO2) and measuring in cylindrical gas proportional counters the radioactivity through the beta particles which are emitted when the 14C atoms decay8. The Laboratory was responsible for the processing and measurement of 53 samples, all from Greek sites (code number DEM-).

  • The Centre for Radiocarbon Dating at Lyon (University Claude Bernard Lyon 1, CNRS and University Lumière Lyon 2), which uses the radiometric Liquid Scintillation Counting technique (LSC). In this method, the sample is converted into liquid (benzene) and the radioactivity is measured in liquid scintillation counters by the light released (scintillations) when the beta particles, emitted by the 14C atoms decaying, ionize a special scintillation medium mixed with the sample. The Laboratory was responsible for the processing and measurement of 13 samples from Bulgarian sites (code number Ly-).

  • The Laboratory of Carbon-14 Measurement at the Center of Atomic Energy at Saclay (LMC14), which measured 126 samples from both Greece and Bulgaria with the Accelerator Mass Spectrometry (AMS) technique. Here, the samples are converted into solid (graphite). By bombarding them with cesium ions, the carbon atoms are extracted and accelerated in a tandem accelerator, then separated from the other isotopes according to their mass by strong magnets, and measured directly. The preparation of these samples was also made at Lyon, which explains their double code number (Lyon‑/SacA-).

  • 9 Olsson 1979; Mook & Streurman 1983.

7All samples underwent an initial careful pre-treatment in order to remove any carbon compounds of non-archaeological origin9, such as soil carbonates, humic acids, etc., which are not related to the tissues of the sample itself and could shift its true age if not removed. Typically a standard acid-base-acid treatment was used (ABA, also sometimes called acid-alkali-acid, AAA). The techniques of sample pretreatment and preparation are based on international protocols, but each laboratory has its own control procedures that also depend on the form in which the samples are converted for measurement. These procedures for each of the laboratories involved in the project are described below.

Procedures at the Laboratory of Archaeometry, NCSR “Demokritos”

Pretreatment of charcoal samples

8After a mechanical cleaning, during which all obvious non-charcoal particles were removed from the sample, and a light grinding of the charcoal to smaller particles, the samples were put into a 4% solution of HCl acid at 80 oC and stirred well for at least 30 minutes and as long as necessary to dilute any carbonates from the soil present. The samples were then transferred to 4% NaOH solution, stirred well, and left overnight at room temperature. Following that, the samples were again put into an acid solution of 4% HCl at 80 oC and stirred for more than an hour. They were then neutralized with de-ionized water and dried in an oven at 90 °C.

Pretreatment of bone samples

9he bone was broken into approximately equal size pieces, about 2 cm long. The spongy bone and any encrustations were removed with a lancet. The samples were then washed with de-ionized water and put in an ultrasonic bath to remove soil or dirt precipitations. Samples were then placed in a 0.6N HCl acid solution at 5 oC and the acid was changed frequently until complete demineralization. This process can take weeks. The soft collagen was then transferred into a base solution (0.5% KOH) at 5 oC, which was changed daily until the solution was clear and then put into acid again (0.6N HCl) overnight. Following that, the samples were neutralized and placed in a copper disk in a drying oven at 90 oC and gelatinized.


  • 10 De Vries & Barendsen 1953; Münnich 1957; Nydal 1983.
  • 11 Kromer & Münnich 1992.

10After drying, all samples were combusted using a de Vries-type continuous combustion system10 and converted to CO2. All the other oxides were removed by reaction with KMnO4 and the CO2 was precipitated to calcium carbonate in a CaCl2/NH4 solution. Consequently, the samples were again turned into CO2 by treatment with HCl acid. In the final purification step, the impurities of the gas were removed by passing the sample through a column filled with activated charcoal kept at 0 oC11. Finally, the mass of every sample was adjusted to a fixed amount and then measured in the counters.

Measuring system

11The 14C measuring system consists of a series of cylindrical gas proportional counters made of pure copper with capacities of 4 and 3 L, bearing in the middle a tungsten gold-plated wire of diameter 20 μm. The counters are surrounded by continuous flow (Ar + 10% CH4) guard counters, which monitor all incoming environmental radiation and separate it electronically from the actual sample counts by an anticoincidence system. The samples are alternated every few days between the different counters and measured repeatedly, ensuring the accuracy and reliability of the results.

  • 12 Ibid.

12Very frequently, background (CO2 containing no 14C) and standard samples are run in the counters and correlations with cosmic radiation and purity of samples are obtained and used in the calculations for high precision12. A small aliquot of each CO2 sample is received in small ampoules before the final mass adjustment and sent to an isotope ratio mass spectrometer for measurement of the δ13C value, which is used for fractionation correction purposes (see infra).

Procedures at the Centre for Radiocarbon Dating at Lyon and the LMC14 at Saclay

13As already said, some 140 samples from Greek and Bulgarian sites were processed at the Radiocarbon Laboratory at Lyon. A dozen of them were measured at Lyon itself using the Liquid Scintillation method (LSC), whereas the others were measured by AMS at Saclay. The pretreatment procedures are essentially the same in both cases; only the duration of exposure to the reagents differs (the bigger the sample, the longer the reaction time).

Pretreatment of non-carbonate samples (charcoal, charred fruits or seeds, burnt bones)

  • 13 Goh & Molloy 1972.

14The non-carbonate materials underwent a standard acid-base-acid pretreatment: it consisted of an initial wash in hydrochloric acid (2N), followed by washing with sodium hydroxide (0.1N) and a final acid wash. The purpose was to eliminate the sedimentary carbonates and other organic contaminants, acidic pollutants such as humic and fulvic acids, as well as carbon dioxide from the atmosphere that may have been absorbed during the wash in the base solution13. The duration and intensity of the treatment varied depending on the fragility of the material. Each acid or base washing was followed by rinsing with ultra-pure water. After each rinse, the samples were separated by decantation or centrifugation. After the final rinse, the samples were dried in an oven at 100 °C prior to combustion.

15The sample pretreatment was done manually in glass containers, all of which were heated at 450 °C for a minimum of 5 hours to remove organic contaminants before being used. The samples were handled in lots of 15, composed by identical or similar materials requiring the same treatment. Standard samples of known age (modern and background) were treated together with the archaeological ones, in order to ensure the quality of the procedure, and also to provide data for age calculations.

Pretreatment of bone and teeth samples

  • 14 See Bocherens et al. 2005.

16The final sample size depends on the precise type of material, the available amount, and the degree of conservation of the organic matter. When the bones seemed badly preserved, measurements of nitrogen and carbon in the raw bones were conducted to determine if dating was possible. The bones were processed only when the nitrogen content was > 0.4%14. The determination of carbon allowed us to identify possible pollutants and adapt the preparation: when the preliminary analysis indicated high levels of excess carbon, the collagen was treated by ultrafiltration that ensures a very high purity of the samples, and measured by AMS.

  • 15 Longin 1971.

17The surface of bones and teeth was initially cleaned by abrasion. During sampling, we avoided areas that had been consolidated or parts to be preserved, and we also eliminated all the spongy material. The sample was then crushed and ground finely. Pretreatment of bone involves a simple ABA treatment followed by extraction of collagen15. Thus, the ground samples were treated successively with hydrochloric acid (1N), sodium hydroxide (0.1N), and hydrochloric acid (1N), with rinsing by centrifugation with ultrapure water to bring the pH to 7 between each reagent. Raw collagen was extracted in a solution at pH~3 to ~100 °C for 20 hours. The solution containing the collagen was then filtered on Whatman or Nagel paper to remove the pellet and then lyophilized.

  • 16 At Demokritos we did not proceed to this kind of measurement.
  • 17 See infra, chapters 5 and 6.

18When the amount of collagen derived from the original sample is less than 1% of the weight of the bone material used, we generally consider the dating less reliable from a physico-chemical point of view – which does not mean though that the result of the measurement will not be valid16. This was indeed proved in the case of “Balkans 4000”: the 10 samples that fell into this category (table 1: samples having a collagen content lower than 10 mg/g) provided results that are perfectly compatible with those taken from more reliable samples from the same contexts (see for instance the two samples from grave 1 at Akladi Tseïri: Lyon‑6027 and Lyon‑7484, or the samples D11 and D12 from Kastri: Lyon‑7915 and Lyon‑7914). The case of Borovan is somewhat more questionable, because all three samples were poor in collagen and the results did not correspond exactly to the expected chronology: yet, they fit well the general regional sequence as it is established from neighbouring sites with more abundant and more reliable series (Bezhanovo), and should therefore be considered as reliable as well17.

Combustion and measurement

19The samples for liquid scintillation counting (LSC) were introduced into a combustion bomb and burned in the presence of oxygen. The produced carbon dioxide (CO2) was then chemically transformed into acetylene and benzene. All steps, from burning to the acquisition of benzene, were conducted under vacuum to avoid the introduction of atmospheric carbon dioxide containing 14C. The obtained benzene was weighed and mixed in a glittering product that can detect 14C atoms decaying over time, and the sample was then put for counting: they spent between 1000 and 2000 minutes in each of the six counters of the laboratory.

  • 18 This is why we chose not to give them at all in table 1, in order to avoid misuse.
  • 19 Cottereau et al. 2007. The French National Centre for Scientific Research (CNRS) is one of the inst (...)

20For AMS counting, a few miligrams of the sample in solid form were placed in clean tin capsules and weighed. They were then combusted in an elemental analyzer NC (Thermo Finnigan Flash EA 1112) under a helium stream in the presence of copper oxide at 980 °C. The CO2 produced in the combustion was transferred to a glass ampoule, which was then sealed and sent to the Laboratory for Carbon 14 Measurement at Saclay. There the CO2 gas was reduced to solid graphite and deposited on a steel lamina by heating in a reducing atmosphere. This graphitization process introduces significant isotopic fractionation. Thus the δ13C values obtained from the graphitized samples, although still perfectly usable for the correction of the 14C ages, are considerably different from those of the original samples and cannot be used for other purposes18. The graphite samples were placed in the accelerator. Then the carbon atoms were extracted by cesium ion bombardment and accelerated by a high positive electric potential. Finally the carbon isotopes were separated by strong magnetic fields and collected at suitable detectors. The Accelerator Mass Spectrometer used is a NEC tandem accelerator of 3MV, labeled “Artemis”19.

21The rigorous sample preparation and measurement procedures of all laboratories involved in the project ensured high accuracy and reliability of the dating results. All three laboratories participate in the International Radiocarbon Intercomparison programs, which ascertain the suitability of their facilities and the correctness of the procedures followed. In addition, in several instances during the “Balkans 4000” project samples from the same strata were measured in parallel for cross-checking of the results.

Calculation of the radiocarbon age and calibration

  • 20 See Evin & Oberlin 2005, p. 88; Maniatis 2013, p. 4‑5.

22Irrespective of the method used for measuring the 14C concentrations, the radiocarbon ages were calculated using the standard international conventions20. Thus the 192 radiocarbon ages produced, expressed in years BP (Before Present = 1950 AD), can be directly compared with those produced by other laboratories in the world. The errors which accompany the dates include statistical errors from the measurement and all other laboratory errors and are reported next to the age as “±” values. A final correction to the calculated age for sample isotopic fractionation was performed using the difference of the δ13C value of the sample, measured from its 13C/12C ratio normalized to the same ratio of the international standard of PDB, from the conventional value δ13C (= -25.0‰) of wood.

23The radiocarbon ages were all calibrated at the same degree of confidence (95.4%) using the latest calibration curves (infra). Their results are expressed here in calendar years (cal BC), listed next to the original BP radiocarbon ages. In order to understand completely the implications of this conversion – actually, this correction – it might be useful to recall briefly some of the calibration principles.

  • 21 The long-term variations are due to the variation of the magnetic field of the earth, while the sho (...)

24Radiocarbon ages are not true calendar dates, because of the conventions used for their calculation, namely the use of the old Libby’s half-life for the 14C isotope (5568 years) and, more importantly, the use of a fixed value (the concentration of 14C in an oxalic acid standard) as the initial concentration when the organism was alive (modern value), assuming a constant 14C concentration in the atmosphere globally and in all periods. But the concentration of 14C in the atmosphere has never been constant: there is an overall long-time increase as one goes back in time up to about 30,000 years, and on top of that there are several small short-time variations (wiggles)21. Because of these variations in the 14C concentration in the atmosphere over time, different organisms (plants, animals, and humans) have a different initial concentration of 14C in their tissues depending on the period in which they lived. A radiocarbon age calculated using the modern standard value as the initial 14C for every sample irrespective of age, a procedure necessary for the dates from all laboratories to be directly comparable, would show therefore smaller or bigger shifts from the true age.

  • 22 See Stuiver & Long 1993.
  • 23 Reimer et al. 2009.

25In order to take these variations into account, calibration of the radiocarbon dates is necessary22. This is achieved through a calibration curve, which is constructed using samples of very accurately known age (tree rings, corals, lake varves, etc.) and measuring their radiocarbon age. The calibration curve correlates the radiocarbon age (vertical axis) with the true calendar age (horizontal axis) for a continuous series of known age samples. It starts from the year AD 1950 and goes back in time, recently having been extended to 50,000 years, the limit of the application of the technique23. Here is shown the section of the curve for the last 8000 years, i.e. from the Neolithic onwards (fig. 1). In order to calibrate a given radiocarbon date (fig. 2), the standard error (± 30 years BP in this case) is augmented by 3 fold and a Gaussian (normal) distribution is formed around the central age (5115 BP). This distribution is then projected horizontally on the part of the calibration curve corresponding to the specific period and its intercept with the curve is projected down vertically on the true calendar age axis. In this way the radiocarbon age in BP together with its error is converted to a calendar age range (BC or AD).

  • 24 Reimer et al. 2004.
  • 25 Reimer et al. 2009.
  • 26 A new version was released in 2013, when the present volume was in the final stage of preparation. (...)

26A major upgrade occurred for the calibration curve in the year 2004. The dataset for terrestrial samples produced then is called IntCal0424. This upgrade involved corrections of some flaws that existed in older versions due to incorrect counting of tree rings; this affected some periods to a smaller or larger degree. The curve also extended to about 22,000 years, from the earlier versions which stopped at about 12,000 years. In 2009 a new curve became available with a dataset for terrestrial samples known as IntCal0925. No major corrections were applied to the 2009 dataset with respect to the 2004 version but the new curve was extended to the full range of application of the technique (close to 50,000 years)26.

27In fact, calibration allows doing more than that: combined with statistic analysis, it allows passing from single values with a uniform margin of error (e.g. 5115 ± 30 BP) to time intervals (e.g. 3978‑3801 cal BC), with an internal distribution of probabilities; specific dates within this range can have a higher or lower probability to be the true age of the event (fig. 2). The overall extent of the interval is shorter or longer according to the natural variations of the curve (wiggles, “plateaux”) at a particular point, but also according to the deviation applied to the original radiocarbon age before the projection on the calibration curve. This deviation can be of 1 sigma, when the original margin of error of the radiocarbon age is augmented only once, or 2 sigmas, when the margin of error is augmented twice before projection. The simple numerical expression of the overall range of probabilities (68.2% in the case of one sigma, or 95.4% in the case of two sigmas), does not allow one to see immediately their internal variations, which are however of great importance if one wants to proceed to further statistical analysis of results. These variations can be expressed numerically (in lists), but are more usually presented as higher or lower peaks on graphics (like the one shown here, fig. 2). The final calibrated age ranges take also into account the standard error of the curve (width of the curve in fig. 2).

Fig. 1 – The calibration curve in the last 8000 years. The radiocarbon age in BP is plotted on the vertical axis versus the calibrated age in calendar years (BC or AD) [Reimer et al. 2009].

Fig. 2 – Typical age calibration of a sample (DEM-2095) performed with the calibration program OxCal v4.2.3 using the IntCal09 dataset (Reimer et al. 2009).

  • 27 For instance Görsdorf & Boyadzhiev 1996.
  • 28 See also Maniatis 2013, p. 13‑15.

28In our study we chose to apply the 2s deviation instead of the 1s sometimes preferred by others27, in order to maximize the probability that the dates of the events under investigation fall within the suggested interval of calendar years. Indeed, although calibration at 1s gives, as we said, ranges that are shorter and therefore look more precise, it presents a certain risk (more than 31%) for the true age of our events to be older or younger than the range given. With the 2s interval, our results are less precise but more secure, since there is a less than 5% probability that the true age stands outside the suggested limits28.

  • 29 See Boyadzhiev 1998, p. 349.

29There have been some doubts expressed as to whether it is preferable to discuss archaeological sequences using calibrated dates, or if it would be better to use them uncalibrated, knowing of course that the true ages could be off by about 1000 years or more29. In the case that interests us here, the question is almost secondary, for the patterns presumably observed in the radiocarbon dates of the Balkan Chalcolithic (gaps in occupation, chronological differences between some sites or areas: supra, chapter 1) describe relationships between sets of data, and therefore are seen both in the calibrated and the uncalibrated results. On the other hand, the argument that calibrated dates can hide differences between individual events is true, especially in cases where the calibration curve shows very close and sharp variations, resulting in a “plateau” (like the one known, for instance, for the years between 4000‑3800 cal BC or the years 3300‑3000 cal BC, to take only two examples within the limits under consideration). But the reverse is also true, i.e. that different radiocarbon ages can correspond to events of the same true age: in this case, putting them into an absolute order according to their main BP value can be misleading (especially if they have great variations in the margin of error) and calibration within the same probability range is the only way to reveal their hidden affinity.

30Table 1 (p. 50‑65) provides the results of the 192 dates conducted in the frame of the “Balkans 4000” project. We have indicated the datasets used for the calibration (IntCal04 or IntCal09) just for historical reasons, as there is absolutely no difference between these two curves.

Global evaluation of the results

31The historical consequences of the results retrieved from the radiocarbon measurements conducted in the frame of the “Balkans 4000” project are discussed in detail in the following chapters. Here we simply consider them from the point of view of procedure.

32The radiocarbon dates produced, as a rule, fit perfectly the expected relative chronology both broadly and in the details. Indeed, when a stratigraphy exists, the dates are ranged accordingly; this proves that the dating procedure has been correct all the way along, from the “reading” of the contexts and the selection of samples down to the measurements.

  • 30 In addition of course to that of contamination or poor preservation of the organic material, which (...)
  • 31 See infra, chapter 17, p. 323, n. 30 and p. 336, n. 109. The values given here, table 1, are the re (...)
  • 32 See infra, chapter 7, p. 127.

33In the very few cases where this does not happen, i.e. where the obtained results do not correspond to the expected chronology (dates shown in italics in table 1), several explanations can be considered30. The first one concerns the mixing of samples during dispatching or processing. This was the case of two samples from Sidirokastro, where inversion at the laboratory has been proven by remeasuring a new sample extracted from the original bag31. In the case of four samples from the first series from Karanovo, inversion at the laboratory has been excluded, but resampling and measurement of a second series of samples suggests that the discrepancies of the first series were due to mixing during packaging and dispatching, or to taphonomical problems gone unnoticed at the excavation32.

  • 33 Evidence from material suggests however that the site was occupied at least during the first centur (...)

34The case of Sadovets-Ezero is probably different; indeed, we seem to deal here with an overall merging of dates from samples that were assigned originally by the excavator to three successive horizons. The 14C dates suggest that either the tripartite division of the site’s Chalcolithic deposits is altogether exaggerated, and therefore the site’s internal stratigraphy should be simplified, or the distinction is right but the selected samples do not represent it correctly. Naturally, the answer to this question requires further analysis by the archaeologist in charge. All we can say at present is that, if the end of the settlement is indeed represented among the dated samples, it does not go beyond the last quarter of the 5th millennium BC33.

  • 34 See infra, chapter 3, p. 81‑83.
  • 35 Supra, chapter 1, p. 15‑17.
  • 36 Boyadzhiev 1995, p. 171; Boyadzhiev 1998, p. 367, n. 8.
  • 37 Supra, chapter 1, p. 33.

35In the cemetery of Smyadovo, we had a horizontal stratigraphy, with graves that looked definitely Late Chalcolithic on one hand and graves that looked definitely EBA on the other hand, together with a few uncertain specimens in between. The excavator postulated that these atypical graves would fall, precisely, in the transitional period between the two others. But the 14C dates suggest a different explanation: instead of representing a different chronological phase, these graves may represent rather different burial customs within those same periods34. This discovery, whose historical consequences are important for our understanding of the changes that take place in the Balkans already during the second half of the 5th millennium BC, adds to the discussion about the relative weight of the evidence from material culture and about the risks of taking “cultures” and “phases” as strictly synonymous35. This discussion should not be confused with the one about the presumed “ageing” of some of the Chalcolithic dates from Smyadovo and from other sites in Bulgaria36. As has already been pointed out, the claim that they may reflect local anomalies in specific parts of the calibration curve is solely based on archaeological considerations about how things should be (what would be the “normal” thickness of occupation layers, or the expected duration of specific phases), and is not justified from scientific evidence properly speaking37.

  • 38 Supra, chapter 1, p. 25 and 34.

36Serviana is an exception in the general picture, and at the same time a good example of the kind of answers that 14C dating can provide in cases of archaeological contexts of poor diagnostic character. The site has been tentatively attributed to the transitional period between the Neolithic and the EBA, based on its overall setting, which recalls vaguely the topographical situation of the neighbouring Doliana38, and on the recovered pottery, which was coarse and undecorated. But radiocarbon dating has shown that the site (at least the dated feature) belongs to another “transitional” period, the Early Iron Age.

37We will not comment upon all the examples of pits troubling the reading of the stratigraphy, either because their limits were unclear (case of Kosharna), or because they were filled again with material from the surrounding, i.e. earlier, layer (case of Dikili Tash). As we said, we tried to avoid them as much as possible and were extremely careful every time we actually took samples from pits, but by all evidence this is never enough. What we should probably keep in mind as a general “lesson” is that, every time we move away from a floor deposit or an undisturbed destruction layer, we risk mixing or misunderstanding of the samples’ real context. Things like that probably happen more often than we think, and may be responsible for many of the observed “anomalies” in the chronological record.

Table 1 – Radiocarbon dates conducted in the frame of the “Balkans 4000” project.

Table 1 – Radiocarbon dates conducted in the frame of the “Balkans 4000” project.

* expected relative chronology at the time of the samples’ submission; given according to the terminology used by the excavators.


4 Supra, chapter 1, sp. p. 36‑37, fig. 2, and table 2.

5 See McFadgen 1982; Ashmore 1999.

6 Supra, chapter 1, p. 32.

7 See also Evin & Oberlin 2005, p. 84‑88; Maniatis 2013, p. 6.

8 Maniatis 2013, p. 7‑10.

9 Olsson 1979; Mook & Streurman 1983.

10 De Vries & Barendsen 1953; Münnich 1957; Nydal 1983.

11 Kromer & Münnich 1992.

12 Ibid.

13 Goh & Molloy 1972.

14 See Bocherens et al. 2005.

15 Longin 1971.

16 At Demokritos we did not proceed to this kind of measurement.

17 See infra, chapters 5 and 6.

18 This is why we chose not to give them at all in table 1, in order to avoid misuse.

19 Cottereau et al. 2007. The French National Centre for Scientific Research (CNRS) is one of the institutions who benefit from special financial conditions for access to this accelerator. It is thanks to this privileged condition that it has been possible to process so many dates with the AMS technique, otherwise very expensive.

20 See Evin & Oberlin 2005, p. 88; Maniatis 2013, p. 4‑5.

21 The long-term variations are due to the variation of the magnetic field of the earth, while the short-term ones to sun flares or quiet periods. A strong magnetic field of the earth and an active sun means scattering of the cosmic radiation and hence less production of 14C, whereas a weak field and a quiet sun mean more production. See among others Kitagawa & Van der Plicht 1998.

22 See Stuiver & Long 1993.

23 Reimer et al. 2009.

24 Reimer et al. 2004.

25 Reimer et al. 2009.

26 A new version was released in 2013, when the present volume was in the final stage of preparation. No changes were imposed in the dataset but only a slightly better shaping of the curve using modern statistical analysis. The effect on the calibrated ages using the 2013 curve are negligible and have no impact whatsoever on the results of this work.

27 For instance Görsdorf & Boyadzhiev 1996.

28 See also Maniatis 2013, p. 13‑15.

29 See Boyadzhiev 1998, p. 349.

30 In addition of course to that of contamination or poor preservation of the organic material, which are the first to come to mind. Yet, no case of true contamination has been detected among the “Balkans 4000” series (i.e. samples giving an aberrant, inexplicable result). The case of bone samples with low collagen content has been commented on supra.

31 See infra, chapter 17, p. 323, n. 30 and p. 336, n. 109. The values given here, table 1, are the reconstituted, right ones.

32 See infra, chapter 7, p. 127.

33 Evidence from material suggests however that the site was occupied at least during the first centuries of the 4th millennium BC, which means that the samples labelled “phase VIII” probably did not actually represent this horizon. For the pottery affinities from the site, see also the comment of M. Valentinova, infra, chapter 5, p. 112.

34 See infra, chapter 3, p. 81‑83.

35 Supra, chapter 1, p. 15‑17.

36 Boyadzhiev 1995, p. 171; Boyadzhiev 1998, p. 367, n. 8.

37 Supra, chapter 1, p. 33.

38 Supra, chapter 1, p. 25 and 34.

Table des illustrations

Légende Fig. 1 – The calibration curve in the last 8000 years. The radiocarbon age in BP is plotted on the vertical axis versus the calibrated age in calendar years (BC or AD) [Reimer et al. 2009].
Fichier image/jpeg, 148k
Légende Fig. 2 – Typical age calibration of a sample (DEM-2095) performed with the calibration program OxCal v4.2.3 using the IntCal09 dataset (Reimer et al. 2009).
Fichier image/jpeg, 128k
Titre Table 1 – Radiocarbon dates conducted in the frame of the “Balkans 4000” project.
Légende * expected relative chronology at the time of the samples’ submission; given according to the terminology used by the excavators.
Fichier image/jpeg, 440k
Fichier image/jpeg, 524k
Fichier image/jpeg, 560k
Fichier image/jpeg, 532k
Fichier image/jpeg, 456k
Fichier image/jpeg, 496k
Fichier image/jpeg, 504k
Fichier image/jpeg, 508k
Fichier image/jpeg, 480k
Fichier image/jpeg, 516k
Fichier image/jpeg, 512k
Fichier image/jpeg, 524k
Fichier image/jpeg, 500k
Fichier image/jpeg, 528k
Fichier image/jpeg, 480k
Fichier image/jpeg, 354k
Cette publication numérique est issue d’un traitement automatique par reconnaissance optique de caractères.
Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search