14 C dates and stratigraphy > reconsidering the sequences at Moverna vas ( Bela Krajina , southeastern Slovenia )

The vertically stratified sites have always been important in conceptualisations of prehistoric cultures and time. Time is conceptualised through the building of stratigraphic matrices, which allow the recognition of the chronological sequence of archaeological deposits and find assemblages. What stratigraphic matrix lacks, however, is temporal depth, the span of calendric time in which deposits and assemblages are formed (Lucas 2012.121).


Introduction
The vertically stratified sites have always been important in conceptualisations of prehistoric cultures and time.Time is conceptualised through the building of stratigraphic matrices, which allow the recognition of the chronological sequence of archaeological deposits and find assemblages.What stratigraphic matrix lacks, however, is temporal depth, the span of calendric time in which deposits and assemblages are formed (Lucas 2012.121).
The 14 C method allows the dating of deposits and assemblages within the calendar time-frame.The Bayesian approach to building calendar chronologies (Bronk Ramsey 2009;Bayliss 2009) is especially useful for temporalising the stratigraphic matrix, as it provides formal probabilistic estimates for the ages of events important to archaeologists or events that might have actually been experienced by people in the past.Furthermore, it provides the duration of activities recognised in the archaeological record.This enables us to exceed the stratigraphic conceptualisation of time as a sequence and consider the tempo-rality of formation processes that created archaeological deposits and assemblages at sites.A much less discussed utility of 14 C dating is the ability to recognise and rectify the problematic parts of the stratigraphic matrix.In establishing chronological sequences, the 14 C dating and the stratigraphic matrix should be used as complementary methods, especially in cases where the stratigraphic relationships are harder to observe and where the archaeological deposits are finer than the practice of stratigraphic excavation and recording is able to discern.The Bayesian approach is useful here as a tool for testing hypotheses and alternative chronological scenarios.
In the paper, we present the case study of the Moverna vas site calendar chronology (Budja 1994), where recently obtained AMS 14 C dates contradict the stratigraphic matrix in its uppermost part.By confronting the two sets of data, we suggest a partial adjustment of the stratigraphic matrix, with the repositioning of some stratigraphic units.Furthermore, the 14 C dates reveal two groups of probability ABSTRACT -Recently obtained AMS 14 C dates and the stratigraphic matrix from the Moverna vas site show inconsistencies in the upper part of the stratigraphic sequence.The Bayesian approach to chronological modeling in the OxCal program is used in the paper to present the calendar chronology of Neolithic and Eneolithic settlement phases at the site and to propose a revision of the upper part of the stratigraphic sequence, also partly supported by a typological comparison of a pottery vessel from one of the 14 C dated contexts.
The Moverna vas site is located in the Bela Krajina Region (southeastern Slovenia) on a plateau overlooking and bordering on three sides the bending gulch that the River Krupa has carved through an undulating lowland Dinaric karst landscape (Fig. 1).The observations and analyses conducted at the site provide us with a long stratigraphic and typological sequence of deposits and find assemblages spanning the 5 th and 4 th millennia calBC (Budja 1988;1989;1990a;1990b;1992;1994;1995;Andri≠ 1993;Toma∫ 1997;1999;Ωibrat Ga∏pari≠ 2008;Sraka 2012).
A series of excavations was carried out at the Moverna vas site in the 1980s, with excavations in 1988 being one of the first in Slovenia that complied with the principles of archaeological stratigraphy (Budja 1988;1990a;1994).The stratigraphic matrix (Fig. 2) shows the superposition of naturally formed layers in which finds were deposited more or less accidentally, indicating indirect traces of settlement.These layers were largely deposited over the whole excavation area, were vertically separated by recognised interfaces and contained more direct remains of human life at the site, such as post-holes, refuse pits, hearths and collapsed daub walls of houses interspersed within the deposits and on the former living surfaces.According to the excavator, "The stratigraphic sequence in Moverna vas shows a repetition of natural processes and anthropogenic activity.We recognized the former as processes of loam and soil sedimentation and erosion and the latter as building, destruction and everyday activity [translated by the author]" (Budja 1990a.127).
The linear stratigraphic sequence of layered deposits was integrated with the interlinear correlation of various stratigraphic units, such as posts, pits and walls, which were supposedly associated with human activity and can be considered roughly chronologically synchronous (Budja 1990a.127;1994.18-21).The integration of the two lead to the conceptualisation of the temporally structured archaeological record at Moverna vas as a sequence of cultural phases of human occupation or settlement phases.The sequence of phases is represented by repetitions of activities such as building, destruction and everyday life, complemented by continuous natural pro-cesses of erosion and sedimentation (Figs. 2 and  3).Nine settlement phases were recognised, with phases 2 through 6 being associated with the Neolithic period and phases 7 through 9 with the Eneolithic period, according to the pottery typology.The supposed chronological association of find assemblages was examined with the evaluation of the dispersion of pottery fragments that once formed the same vessel within the individual stratigraphic units.Because the vertical dispersion of fragments is delimited by the layer interfaces, the find assemblages consistently represent individual settlement phases at the site (Budja 1990a.130-2;1994.21).

C samples and their stratigraphic contexts
Thirty-seven 14 C dates are currently available from the Moverna vas site (see Appendix), of which twenty-seven were obtained from carbonised residues on pottery and the remainder from charcoal samples.The first charcoal samples were obtained from metrically defined layers of the un-stratigraphic excavations in 1984 (Budja 1988.50-51;1989.97)and were measured at the Ru∂er Bo∏kovi≤ Institute in Zagreb by the gas proportional counting technique (Srdo≤ et al. 1987.139).Further charcoal samples were obtained from stratigraphic layers related to the Neolithic phases after excavations in 1988 and were dated at the Oxford Radiocarbon Laboratory (Budja 1994.Fig. 5).A programme of direct AMS dating of carbonised residues from the interior surfaces of pottery vessels was recently initiated, together with the application of lipid extraction and characterisation of the organic residues absorbed in vessel walls.Samples relate to all phases and were dated at the Poznan Radiocarbon Laboratory (Ωibrat Ga∏-pari≠ 2008.Fig. 5.1;Sraka 2012.appendix;see Appendix).

Fig. 1. Location of Moverna vas site in the Bela Krajina region, southeastern Slovenia.
In the latest dating programme, 14 C samples were obtained from the previously undated Eneolithic phases, in order to expand the chronological model and sketch out the general chronological structure of the whole stratigraphic sequence.Short-lived samples in the form of carbonised residues on pottery were preferred, but due to their limited preservation in the Eneolithic assemblages, mostly charcoal was sampled.All 14 C dates from the site are listed in the Appendix, together with the relevant contextual information.In Bayesian modeling, only dates on samples from the stratigraphic excavation in 1988 are used.
Figure 3 shows the modified composite section of stratigraphic units associated within individual phases.New 14 C dates originate from each of the Eneolithic layers (009.2, 009.1, 006) as well as from a patch of burnt daub (020) and refuse pits (033,046,030,011).Each layer corresponds to originally defined phases 7, 8 and 9, respectively, with pits and the daub patch also associated with separate phases.
In all cases, the dates from pits (but not the one from the daub layer) contradict those from the layers.Dates from pits 033, 046 and 030 have much younger calendar ages than those from layers 009.2 and 009.1, with which they are integrated in the 7 th and 8 th phases, respectively.The probability distributions of dates from pits 033 and 030 (Poz-54005, Poz-54007) partly overlap with the 14 C date on carbonised residue from layered deposit 006, associated with the 9 th phase (Poz-53998).The probability distribution of dates from pits 046 and 011 (Poz-54003, Poz-54009) lie somewhere between the 14 C dates from layers representing the 8 th and the 9 th phase.The probability distributions of the problematic 14 C dates thus fall into two groups, one before the middle of the 4 th millennium and the second around the turn of the 4 th and 3 rd millennium calBC (Fig. 4b).These dates contradict the established stra-  Budja 1994.Fig.7).

Fig. 3. The composite section, showing associated remains of human occupation within individual phases and the indication of the revised parts of the stratigraphic sequence (adapted after
tigraphic matrix, but are invaluable in revealing the problematic parts of the site stratigraphy.
The 'old wood effect' on the charcoal samples, although possible, can not explain the contradiction, as the dates appear younger, not older, than expected.However, sample infiltration is possible, so further, preferably short-lived, 14 C samples are needed to resolve the issue.The unexpected results prompted us to revisit the 14 C sample contexts and their position within the stratigraphic matrix at the site.The stratigraphic matrix can be partly changed with the repositioning of the mentioned pits.The two mentioned groups of probability distributions suggest at least two phases of occupation within the deposit originally associated with the 9 th phase.This is further elaborated in the following section.The suggested modification of the stratigraphic sequence is indicated in Figures 2 and 3.
The inconsistency between the 14 C dates for layers and pits point to the difficulties that archaeologists face when trying to define the level from which pits are cut.All Eneolithic phases at Moverna vas are represented by approx.0.5m of soil deposit on average.So it is not surprising that some of the pits became associated with layers with which they originally were not.Besides the 14 C data, the association of pits 033 and 030 with the originally defined 9 th phase rather than the 7 th or 8 th phase is suggested also by the conservatively drawn cross-section from the un-stratigraphic excavations in 1984 (compared to the rather interpretative cross-section from the stratigraphic excavations in 1988) in the section where the two excavation areas meet.The 14 C dates suggest at least one phase of occupation at the site for which no associated archaeological deposit was recorded, but which is evidenced by pits 046 and 011, previously associated with the 7 th and 9 th phases, respectively (Fig. 3).

Bayesian modeling of the site calendar chronology
The basic idea behind the Bayesian modeling of archaeological chronologies is encapsulated in a simple theorem published by clergyman Reverend Thomas Bayes in the mid-18 th century: posterior = likelihood x prior.This simply means that we analyse the 14 C data we have collected about a problem (likelihood) in the context of our existing archaeological information (prior) in order to arrive at a new understanding (posterior).Put differently, the Bayesian approach integrates 14 C dates and other chronological-ly relevant information available to archaeologists and allows the development of calendar chronologies for a wide variety of archaeological situations, from site-based sequences to regional spatio-temporal phenomena.Despite the fact that the statistical procedures and computing necessary for its implementation are incomprehensible to most archaeologists, computer programs such as OxCal provide simple tools for developing Bayesian models (Bronk Ramsey 1995;1998;2008;2009;Bayliss 2007;2009).The Bayesian approach does not provide final answers and is a heuristic tool for testing and comparing chronological models; the results are never absolute and final and are subject to change when additional 14 C data and archaeological information become available.
In developing a Bayesian model of the Moverna vas site chronology we use the OxCal computer program, version 4.2.3 (Bronk Ramsey 2009) with the implemented IntCal13 calibration curve (Reimer et al. 2013).The intention is to use the sequence of phases, explained above and shown in Figures 2 and 3, as prior archaeological information for grouping and sequencing the 14 C dates (Informative prior in Bayesian modeling).Specific ways of chronological modeling imply specific assumptions that underlie the statistical procedures implemented in the OxCal program.In our case, the assumption is that each phase represents a delimited period of occupation at the site.
The recent 14 C dating results inclined us to partly change the approach to the Bayesian modeling of the Eneolithic settlement phases, unlike the modeling of the Neolithic phases, which remains unchanged (Sraka 2012.355-358, Fig. 2).The chronological model consists of two parts, the first represented by phases 2 to 8 and the second by phases 8, 9a and 9b.These two parts of the model should be viewed differently in terms of conservativeness and the certainty of the dating results.In the first part of the model, the phases are modeled as contiguous, sharing the same transitional boundary.In the second part, the phases are modeled as sequential, as the preliminary assessment of the 14 C dates suggests temporal gaps between them.The different parts of the model differ in the way that they deal with samples of different quality (short-lived versus long-lived).
In the first part of the OxCal model, the short-lived carbonised residue dates from the pottery vessels are assumed to sufficiently represent human activity related to individual phases.The program assumes them to be uniformly distributed events within individual phases (Uninformative prior in Bayesian modeling).Charcoal dates, on the other hand, are less directly related to activities within individual phases, due to the possible 'old wood effect'.The probability distribution of the two sample types (where both date the same stratigraphic context) do not differ mar-kedly; nevertheless, the 14 C dates on charcoal from the first part of the model are not fully incorporated into the model and are included only as termini post quos (Bayliss et al. 2011.56-58).This means that they only effectively define the ending but not the beginning of phases.Despite the loss of precision, not incorporating the long-lived samples fully into this part of model makes the results more conservative.
In the second part of the OxCal model, however, the distinction between the 14 C dates of different quality cannot be made, as the majority of the samples are on charcoal.Here both carbonised residue and charcoal dates are fully incorporated into the model.The different treatment of parts of the same model may seem inconsistent or even hypocritical.Bayesian modeling is used here as a heuristic tool for testing and gradually improving the models (Bayliss 2007).The 14 C dates which date to the 4 th millennium calBC fall into two clearly separated groups of probability distributions.The first group is represented by 14 C dates from pits 046 and 011, previously associated with the 7 th and 9 th phases, respectively.The second group of probability distributions consists of dates from layer 006, associated with the originally defined 9 th phase, with an addition of the two pits 033 and 030, previously associated with the 7 th and 8 th phases respectively.On the basis of these two groups of probability distributions, the second part of the OxCal model is divided into two sequential phases, 9a and 9b.The second part of the model should be seen as tentative, both because of the lower quality (charcoal dates probably not di-rectly related to the activities and events of deposition) as well as quantity of the 14 C dates, when compared to the well-established first part of the model.

Calendar chronology of the Neolithic and Eneolithic settlement phases on the site
The posterior probability distributions obtained as results of the modeling presented above are presented graphically in Figure 4a for the first part of the model (phases 2 through 8) and in Figure 4b for second part (phases 9a and 9b).The estimated calendar ages for the boundaries between phases and estimated durations of phases are presented in Table 1.The model has an agreement index of 108.5%.One 14 C date on carbonised residue (Poz-21404) from the 6 th phase is inconsistent and is an outlier, as a replicated sample (Poz-48534) on carbonised residuefrom the same vessel shows consistency with the model.One further 14 C date on charcoal (Poz-54008) had to be excluded from the model, as it is significantly later than other dates from the same layer.It is probably an infiltrated sample from later activity on the site.
According to the results, the Moverna vas site was first occupied in the first centuries of the 5 th millennium calBC, while the youngest 14 C dates point to human activity around the end of the 4 th and the beginning of the 3 rd millennium calBC.The span of the calendar chronology of the site is approximately two millennia; however, the site does not seem to have been occupied continuously.The Neolithic part of the model is well established and provides a tighter calendar dating for the phases and their find assemblages.According to the estimated phase durations, phases 4 to 8 follow each other at relatively short intervals of no more than a few human generations per phase and suggest continuous occupation in the Neolithic.The dating of the first two Eneolithic phases, 7 and 8, is less secure, due to the limited number of available 14 C dates and especially due to the plateau shape of the calibration curve in the 2 nd half of the 5 th millennium calBC.The supposed continuity of occupation on the site is inter- rupted in the last centuries of the 5 th millennium calBC.
The latter Eneolithic part of the model with phases 9a and 9b is provisional, and the results are preliminary at best.Phase 9a can probably be dated to the 37 th or 36 th century calBC.Phase 9b can be very coarsely dated to the last centuries of the 4 th and first centuries of the 3 rd millennium calBC.Additional, preferably short-lived 14 C samples are needed from the latter part of the stratigraphic sequence at Moverna vas in order to substantiate the dating of phases 9a and 9b.
The association of pit 030 with phase 9b rather than the 8 th phase is further supported by the typological comparison of an egg-shaped pot found in the pit.(Toma∫ 1999.Pl. MV39.2).The pot is decorated (except for the neck) with barbotine application.The rim is bent backwards and stuck on the outer wall and is decorated with finger impressions.The closest analogies can be sought on the Ljubljansko barje, especially at Parte-I∏≠ica (compare Velu∏≠ek et al. 2000.Pl. 4.3) andParte (compare Harej 1987. Pl. 15.7), where similar backwardly bent rims, decorated with finger impressions and barbotine are numerous.Parte-I∏≠ica is dated to the 29 th and 28 th centuries calBC (∞ufar et al. 2010.2036).The 14 C date on charcoal from pit 030, from which the mentioned pot originates, is actually the youngest of all the dates currently obtained from Moverna vas and its probability distribution partly overlaps with the calendar dating of the Parte-I∏≠ica in the early 29 th century calBC.

Conclusion
Recently obtained AMS 14 C dates contribute to the calendar chronology of the Neolithic and Eneolithic settlement phases at Moverna vas.By confronting the 14 C dates and the stratigraphic matrix, we suggest moving pits 011, 030, 033 and 046 to a different position within the stratigraphic sequence and their re-association with other settlement phases at the site.The 14 C probability distributions suggest a previously unrecognised phase of occupation at the site.The calendar chronology at Moverna vas spans about two millennia of human occupation.In the 5 th millennium, occupation was continuous, with wellestablished calendar dating of archaeological deposits and find assemblages.The discontinuous occupation in the 4 th millennium calBC is at the moment only preliminarily dated.The new calendar chronology allows a more thorough exploration of the palimpsests characterising the archaeological record at Moverna vas and demonstrates the need for 14 C dating to be seen as an integral part in constructing the stratigraphic matrices at sites./ 1999.Pl. MV17-26 ?ibrat Ga[pari;2008.Fig. 5.1

Fig. 2 .
Fig. 2. The stratigraphic matrix and phasing at the Moverna vas site with shaded 14 C dated contexts and the indication of the revised parts of the stratigraphic sequence (adapted afterBudja 1988.Fig.6).

Fig. 4a .
Fig. 4a.Posterior probability distribution of the modeled 14 C dates from phases 2 through 8 in Moverna vas.

Fig. 4b .
Fig. 4b.Posterior probability distribution of the modeled 14 C dates from phases 9a and 9b at Moverna vas.

Calendar age estimates for Boundary events and estimates of durations of Neolithic and Eneolithic phases at Moverna vas.
Tab. 1. .1