UDK: 581.13+591.13:902.65/4-924.56)"633/634' Documenta Praehistorica XXVII Stable Isotopes, Radiocarbon and the Mesolithic-Neolithic Transition in the Iron Gates Clive Bonsall1, Gordon Cook2, Rosemary Lennon3, Douglas Harkness4, Marian Scott5, Laszlo Bartosiewicz6, Kathleen McSweeney7 department of Archaeology, University of Edinburgh, UK, CBonsall@ed.ac.uk 2Scottish Universities Environmental Research Centre, UK, g.cook@surrc.gla.ac.uk 3 Department of Archaeology, University of Edinburgh, UK 4Scottish Universities Environmental Research Centre, UK, d.harkness@surrc.gla.ac.uk 5Department of Statistics, University of Glasgow, UK, marian@stats.gla.ac.uk 6Institute of Archaeological Sciences, Lorand Eotvos University, Hungary, h10459bar@ella.hu 7Department of Archaeology, University of Edinburgh, UK, kath.mcsweeney@virgin.net ABSTRACT - The results of stable carbon and nitrogen-isotope analyses of human bone collagen from the Iron Gates sites of Lepenski Vir, Vlasac and Schela Cladovei are reconsidered in the light of recent developments in stable isotope palaeodietary research and new information on chronology. The revised data have implications for the interpretation of Lepenski Vir and Vlasac, and the timing of the Mesolithic-Neolithic transition in the Iron Gates. IZVLECEK - Vclanku smo preucili rezultate analiz stabilnih izotopov ogljika in dusika iz kolagena cloveskih kosti, ki izvirajo iz najdisc Zeleznih vrat: Lepenski Vir, Vlasac in Schela Cladovei. Pri tem smo upostevali najnovejse izsledke raziskovanja paleoprehrane s stabilnimi izotopi in nove krono-loske podatke. Nanovo pregledanipodatki vplivajo na interpretacijo Lepenskega Vira in Vlasca ter na casovno umestitev mezolitsko-neolitskega prehoda v Zeleznih vratih, KEY WORDS - Iron Gates; stable isotopes; radiocarbon; palaeodiet; Mesolithic; Neolithic; Lepenski Vir; Schela Cladovei; Vlasac INTRODUCTION The Iron Gates has an abundant and continuous record of human occupation in open-air settlements from the Late Mesolithic into the Early Neolithic, c. 8500-6500 BP (7500-5450 cal BC). While early farming settlements are well represented in other parts of the Balkan peninsula, Mesolithic sites are uncommon and there are few if any sites that were inhabited continuously from one period to the next. Therefore, the Iron Gates is arguably the only area of southeast Europe where the transition from Mesolithic to Neolithic can be studied in detail. The evidence for changes in subsistence practices across the Mesolithic-Neolithic transition in the Iron Gates was reviewed by Bonsall et al. {1997). Their assessment was based largely on the results of stable carbon and nitrogen isotope analyses of human bone from Lepenski Vir, Vlasac and Schela Cladovei. The purpose of the present paper is to re-examine the Iron Gates stable isotope data and their significance, in the light of better information on food sources and chronology. DIETARY RECONSTRUCTION FROM STABLE ISOTOPES: SOME BASIC CONSIDERATIONS Stable isotope analysis of carbon and nitrogen in bone collagen has become a standard technique for palaeodietary studies. The underlying principles may be summarized briefly as follows: • stable isotope ratios in bone collagen reflect those in diet • ratios of stable isotopes vary naturally between major food sources • therefore the importance of different foods in human diets can be estimated from the isotopic composition of collagen. Table 1 lists "typical" 813C and 815N values of bone collagen for various major food sources. With each step along the food chain, there is fractionation of one isotope relative to another, resulting in a change in ratio. Most workers assume a slight enrichment in 813C (up to 1%0) and an enrichment in 815N of 3-4%o between the bone collagen of the food source and that of the consumer. Because aquatic food webs are more complex than terrestrial food webs, this results in much higher 815N values at the top of the aquatic food chain. These factors lead to the "expected" values in bone collagen of humans shown in Table 2. Food source 813C (%0) 515N (%<,) C3 terrestrial herbivores -21.0 +5.0 freshwater fish -20.0 +11.0 marine fish -13.0 +13.0 Tab. 1. "Typical" &3C and 8'5N values in bone collagen of three major animal food sources available to humans. The figures for stable isotope ratios in food sources and estimates of trophic level effects cited above should be regarded as "global" averages. There can be significant variation between ecosystems. Therefore, precise dietary reconstruction requires detailed knowledge of the isotopic compositions of local food resources. Humans feeding on: S13C (%<») 815N (%») C3 terrestrial herbivores -20.0 +8.0 freshwater fish -19.0 +14.0 marine fish -12.0 +16.0 Tab. 2. "Expected" &3C and 5,3N values in bone collagen of humans feeding exclusively on each of the food sources listed in Table 1. Other factors need to be taken into account when interpreting stable isotope data. Bone collagen in adult humans is estimated to have a turnover (replacement) rate in a range of 10-30 years (Mays 1998). Therefore, stable isotope ratios are a reflection of average diet over decadal timescales. However, in children the turnover rate is probably more rapid (Kleppinger 1984). Furthermore, the nitrogen isotopic composition of bone collagen is thought to reflect mainly the protein component of the diet as virtually all nitrogen in food comes from protein. Carbon in collagen can be derived from protein, fats or carbohydrates. In high protein diets, the carbon in collagen is thought to come mainly from protein, but in low protein diets a significant proportion of the carbon is probably derived from carbohydrates (Ambrose 1993)- REVIEWING THE IRON GATES DATA Since the original study of the Iron Gates stable isotope data (Bonsall et al. 1997), new information has become available that makes it possible to refine some aspects of the interpretation. This includes information on the isotopic composition of the major food sources, and more accurate age estimates for the human bone samples from Lepenski Vir, Vlasac and Schela Cladovei. There is also a larger data set for Lepenski Vir and Vlasac that can be considered. Food sources In the original study, because of a lack of detailed information on the isotopic composition of local food sources, the human bone stable isotope results from Lepeaski Vir, Vlasac and Schela Cladovei were plotted against data for North American food sources published by Schwarcz {1991), with allowance for fractionation effects (Fig. 1). From this it was concluded that the diets of Mesolithic and Early Neolithic populations were a mixture of foods drawn from two major sources, freshwater fish and terrestrial herbi-vores/C3 plants. It is true that, when compared against the North American data, average 815N values for Mesolithic skeletons from the Iron Gates appear unusually high for a population that subsisted mainly on freshwater fish (cf. Fig. 1). This has led some other researchers (Hedges et al, 1998; Schulting 1999) to infer that the Iron Gates Mesolithic diet must have included a high proportion of Danube-caught anadromous fish from a marine environment, i.e. the Black Sea. The possibility that anadromous fish were the source of the high 815N values was also considered by Bonsall et al. (1997) but was rejected because there was no corresponding enrichment in the 813C values, and because average S15N and 813C values for Mesolithic adults at Schela Cladovei (where there is abundant evidence for Mesolithic exploitation of Fig. 1. S'}C and 5'5N profiles of human populations from Schela Cladovei (A), Vlasac (B) and Lepenski Vir (C) plotted against the ranges of North American aquatic (A), marine (M) and terrestrial (T) food sources, derived from Schwarcz (\ 991 )■ Redrawn from Bonsall et al. (1997). sturgeon) appeared very similar to those of their counterparts at Lepenski Vir and Vlasac, where no sturgeon remains were identified by Bokonyi (1969; 1970; 1977). Although he did not claim to have been a fish expert, masses of large sturgeon remains would not have escaped his attention. Moreover, he did identify Neolithic sturgeon at the site of Mihajlo-vac-Knjepiste (Bokonyi 1992), downstream of Schela Cladovei and the gorge. In this context it is worth noting that research by Ryan etal. (1997) strongly implies that the Black Sea was a freshwater lake until 6700 BP, when there was a rapid influx of salt water as the Mediterranean broke through the Bosporus "dam". If their hypothesis is correct, then all Danube fish exploited during the Mesolithic were freshwater fish, and Neolithic people could not have had access to marine fish until after 6700 BP. This in turn implies that freshwater fish are the source of the very high 815N values recorded in Mesolithic skeletons from the Iron Gates. Existing stable isotope data for aquatic food sources from the Iron Gates are limited to analyses of collagen from three fish bones and an otter bone from Lepenski Vir. These show no consistent pattern, with 813C values varying between -26.3 and -15.7%o and 815N values varying between +8.2 and +12.9%o (.Bonsall et al. 1997). With hindsight, these results may not be unusual. Recent research suggests that 815N and 813C values of freshwater fish can be highly variable (Fig. 2). Moreover, the values can vary quite considerably for different species from the same freshwater system, and for the same species from different freshwater systems. For example, Du- four et al. (1999) report inter-species differences from Lake Constance of ~7%o for 813C and ~8%o for 815N. In the case of the 813C values for the Iron Gates, the spread is almost 11%0. While this spread appears very large, it could be due to a number of factors: ® The number of species in the River Danube could be greater. From the work of Dufour et al. (1999) it can be observed that for each lake under study, the inter-species differences were much greater than the intra-species differences. © The age range of the fish from the study of Dufour et al. (1999) was limited (3-5 years) while those from the Iron Gates could be much greater, as is shown by the evidence of bones from large sturgeon as well as mature carp of extremely large sizes. ® The study by Dufour etal. (1999) was effectively a snap-shot in time, while the samples analyzed from the Iron Gates sites could conceivably span several thousand years and there could have been changes in the freshwater ecosystem within this time-span. ® Most importantly, there could have been a shift to anadromous fish when the Bosporus was breached and the Black Sea became a marine system. The fish specimen with a 813C value of -15.7%o and a 815N value of +12.9%o is certainly not inconsistent with this hypothesis. If a +3-4%o trophic level shift is employed between freshwater fish and human bone collagen (Minaga-wa and Wada 1984), then it would require average 815N values for Danube fish of approximately +10.5%o, or greater, to produce human bone colla- gen values of >+l4%0, which are characteristic of Mesolithic people from the Iron Gates. While this type of data for fish from the Iron Gates is limited, such values are not uncommon. Iacumin etal. (1998) and Pate (1998) report 515N values of about +12%o for Lake Nasser and South Australia fish, respectively. Dufour etal, (1999) report 815N values >+13%o for fish from Lake Geneva and Lake Constance, while Hobson and Welch (1995) report 815N values for large char collected from a high Arctic lake in Canada of >+l4%0. The enrichment of any species will of course depend on the complexity of the food web and its trophic level within the web. Moreover, there is evidence for certain freshwater species that the 815N value increases with the age/size of the fish. This is related to the fact that as a fish grows, it tends to feed at higher trophic levels - in effect, it becomes increasingly carnivorous - and beyond a certain growth stage may start to feed on smaller members of the same species. This phenomenon has been reported for arctic char from Canada (Hobson and Welch 1995). It may also be characteristic of species such as carp, catfish and sturgeon, which dominate fish bone assemblages from the Iron Gates. Many Mesolithic specimens of these species from the Iron Gates sites were very large and, in comparison to their numbers, such fish may have made a disproportionately large contribution to the food supply. 25 20- 4 Freshwater Fish 15- 10- Marine Fish ;T2_...................j Terrestrial Herbivores -5- -40 —rrjTT- -35 -30 -25 -20 I i i i •15 -10 T -5 6I5C (%o) Fig. 2. &3C and S'SN natural variations in freshwater fish (Fl), terrestrial herbivores (Tl) and marine fish from Eurasia Rafter Dufour et al. 1999) compared against North American freshwater fish (F2) and terrestrial herbivores (T2) Rafter Schwarcz 1991) and terrestrial herbivores (T3), river fish (*) and an otter (*) from the Iron Gates fafter Bonsall et al. 1997/ It will be evident from the foregoing discussion and Figure 2 that the North American data on freshwater fish published by Schwarcz (1991) are not necessarily an appropriate model for the Danube, and that locally derived data are to be preferred. While the locally derived data are limited to three fish bone collagen analyses, the isotopic signature for the otter (813C, -19.8%o; 81SN, +10.7%o) effectively provides an average signature for small fish from the Iron Gates, since these will be its primary food source, and arguably, this provides the most reliable model for the Mesolithic human population of the Iron Gates (Fig. 2). An otter's diet comprises primarily small fish, but includes other aquatic animals and small land mammals. Compared to an otter, the bone collagen of a human feeding mainly on much larger fish (and occasionally otters) from the same freshwater ecosystem could be expected to show a slight enrichment in 813C and an enrichment of 3-4%o in 8HN. The data for Mesolithic people from Lepenski Vir, Vlasac and Schela Cladovei are perfectly consistent with this model. While further research is needed into the isotopic composition of aquatic food resources available to Mesolithic and Neolithic peoples in the Iron Gates, new information is available for terrestrial food sources. Collagen values for ungulate bone samples from Mesolithic and Neolithic contexts at Lepenski Vir and Schela Cladovei can be substituted for the North American herbivore data used in the original study (Bonsall et al. 1997). Figure 2 compares the two data sets (T2, T3). Although not all the Iron Gates ungulate samples could be identified to species, probably they derive mainly from deer and cattle. The spread of values on the 813C axis is significantly less for the Iron Gates data set and, while the spread on the 815N axis is similar to the North American sample, the median value (+5.3%o) is significantly higher. Taking into account the various lines of information on the isotopic composition of aquatic and terrestrial food sources, it seems reasonable to continue to use the 815N value as a measure of freshwater versus terrestrial food intake in Iron Gates stone age populations (cf. Cook et al, in press). An end point of +17%o for a 100% aquatic diet is assumed, which is the highest 815N value measured in an adult from the Iron Gates region (Bonsall et al, 1997). For a 100% terrestrial diet a value of +8%o is assumed. This is based on studies by Ogrinc (1999) and Mays (1998) supported by local data for herbivores, as discussed above. Chronology In their original study, Bonsall et al. (1997) published new AMS l4C age measurements on human bones from Lepenski Vir, Vlasac and Schela Cladovei. For all three sites, the human bone ages were older than expected on the basis of existing dates on charcoal - for detailed discussion, see Cook et al. (in press, and forthcoming). From this, it was suggested that the bone collagen of humans who had ingested large quantities of freshwater fish may be depleted in 14C as a consequence of the consumption of material from a reservoir that differed in l4C specific activity from the contemporary atmosphere, thus resulting in l4C ages that are "too old". It was further suggested that this possibility could most easily be tested by comparing radiocarbon age measurements on human bones with those on artefacts of terrestrial animal bone found in the same graves. Schela Cladovei provided material ideal for investigating this problem in the form of bone projectile points, made from long bone splinters of artio-dactyls, found in direct association with skeletons. These were either embedded in human bone or found immediately adjacent to bones of articulated skeletons (which may originally have been embedded in the soft tissue surrounding the bones). In all cases, the bone points may have been the actual cause of death. Cook et al. (in press) obtained AMS 14C dates on paired human and ungulate bone (projectile point) samples. Systematic differences were found between the two sets of ages, demonstrating the existence of a freshwater reservoir effect, and its magnitude was calculated as 540±70 radiocarbon years. From this, and using the 815N value as a measure of the proportion of the diet derived from aquatic foods, it is possible to apply a correction to the human bone ages from all the sites (Cook et al. in press, and forthcoming). The corrected ages are given in Table 3. The effect of the correction is to make the human bone ages significantly younger by approximately 200-500 years depending on the 815N value. However, the reservoir-corrected ages are less precise, i.e. have larger error terms. The expanded data set Bonsall et al. (1997) reported 813C and 815N measurements on 70 individual skeletons - 33 from Lepenski Vir, 29 from Vlasac, and 8 from Schela Cladovei. All the skeletons were those of adults of (supposedly) known chronological context. The Schela Cladovei skeletons were a burial group from below and adjacent to a Mesolithic "house", and seven of them were directly dated by AMS to the Late Mesolithic (Tab. 3). The skeletons from the other two sites had all been recorded as belonging to specific phases of Mesolithic or Early Neolithic occupation -Vlasac I—III ("Mesolithic"), Lepenski Vir I—II ("Mesolithic") and Lepenski Vir III ("Neolithic"). To the original data set can now be added the results from a further 46 skeletons, all from Lepenski and Vlasac. They comprise both adults (25 from Lepen- Site Laboratory Burial 14C Age BP Corrected Calibrated Age ID 14C age BP Range (2o) BC Schela OxA-4384 M52 8570±105 - - OxA-4379 M43 8550±105 8070+122 7450-6645 OxA-4385 M55 8510±105 8090±118 7465-6653 OxA-4382 M49 8490±110 8046+124 7448-6615 OxA-4380 M46 8460±110 8046±123 7448-6640 OxA-4378 M42 8415±100 7971±116 7295-6512 OxA-4381 M48 8400±115 7932+130 7289-6466 OxA-4383 M50 8290±105 7834±120 7061-6439 Vlasac OxA-5824 72 10240+120 9850±130 9949-8843 OxA-5822 51a 8760±110 8380±120 7600-7080 OxA-5827 83 8200+90 7810±105 7049-6441 OxA-5823 54 8170±100 7750±115 7032-6401 OxA-5826 24 8000±100 7600±115 6647-6625 Lepenski Vir OxA-5827 31a 7770±90 7310±108 6404-5926 OxA-5830 44 7590±90 7150±106 6225-5797 OxA-5828 32 7270±90 7040±95 6156-5721 OxA-5831 88 7130±90 6960±93 6011-5644 OxA-5829 35 6910+90 6720±93 5772-5479 Tab. 3. Human bone radiocarbon ages from Lepenski Vir, Vlasac and Schela Cladovei, corrected for the freshwater reservoir effect using method 1 of Cook et al. ^forthcoming). All 14C ages are expressed in conventional radiocarbon years BP (before 1950 AD). The errors are expressed at the one-sigma level of confidence. Calibrated age ranges were determined using CALIB 4.2 fStuiver and Reimer 1993; Stuiver et al. 1998). ski Vir, and 2 from Vlasac) and children (12 from Lepenski Vir, and 7 from Vlasac). The adults are individuals whose chronological context is either unknown or not recorded, or which (in the case of five samples from Lepenski Vir) were assigned to post-Neolithic occupations. These new data are only presented here in graphic form (Figs. 4-7). Full details will be presented in a later publication. All analyses were carried out by the sealed tube combustion method described in Bonsall et al. (1997)• Briefly, this comprises the combustion of small collagen samples in evacuated quartz tubes containing copper oxide as the source of oxygen and a small quantity of silver wire to remove halide contaminants. The CO2 and N2 are then cryogenically separated and analyzed by stable isotope mass spectrometry. The authors consider this to be the most precise and accurate method to determine stable isotope ratios in human bone collagen, and would advise that these and continuous flow measurements should not be combined. In the original study, identification of groups was done primarily by visual inspection of bivariate scat-terplots. In this paper, a variety of exploratory and formal statistical methods, including exploratory cluster analysis, linear discrimination techniques and hypothesis tests have been used to explore groupings and to assess evidence for pre-defined archaeological groups. Statistical analysis was carried out in MINITAB v. 13. DISCUSSION Adults Lepenski Vir The original data set from Lepenski Vir comprised measurements on 33 adults from the various Stone Age occupation phases that were recognized by Srejovic (1969; 1972; Zoffmann 1983) - Proto-Lepen-ski Vir, Lepenski Vir I, II and III. A bivariate scatter-plot of the data (Bonsall et al. 1997; cf. Fig. lc) suggested that there were two groups. One group exhibited 513C and 815N values that were similar to Mesolithic individuals from Schela Cladovei and Vlasac (cf. Figs, la and lb), indicative of diets with a high input of protein from aquatic sources, and the other showed much lower 815N values suggesting diets with increased levels of protein from terrestrial food sources. Cluster analysis of the original data set (Fig. 3) broadly supports these two groupings. The enlarged data set of 58 adults comprises the 33 individuals that had been attributed to Mesolithic and Neolithic contexts, plus 5 individuals "dated" to the Chalcolithic and Medieval periods, and 20 individuals that were not assigned to any occupation "phase". Dietary end points of +8%o and +17.0%o were adopted as representing 100% terrestrial and 100% aquatic diets respectively (Cook etal. in press). Cluster analysis of this revised data set suggests that there are at least three groups, distinguished primarily on the basis of the 815N value (Fig. 4A). The individuals in group 1 have 815N values ranging between +14.4 and +17.0%o, which implies that this group had diets in which 71-100% of the protein was derived from aquatic sources. Group 3 individuals have 515N values between +9.3 and +11.2%o. which implies that the bulk (64-86%) of the protein came from terrestrial sources. Group 2 skeletons have 815N values ranging from +11.8 to +I4.0%o, intermediate between groups 1 and 3, indicating diets in which protein was derived from aquatic and terrestrial sources in similar proportions (42- 67%). The provisionally identified groups (clusters) do not correspond to archaeological (Srejovic) phases and groups contain individuals of diverse age at death and both sexes. However, there appears to be a link between groups and radiocarbon age. Of the five radiocarbon ages currently available (Tab. 3), two lie in group 1 (7310±108 BP, 7150+106 BP), one falls in group 2 (7040+95 BP), and two lie in group 3 (6960+93 BP, 6720+93 BP). The reservoir-corrected ages form a more or less continuous series, and suggest that the three groups relate to different phases in the use of the site. The dendrogram (Fig. 4A) and scatterplot (Fig. 4B) show that sub-groups may exist within the main groups 1 and 3, but their identification is based on only small numbers of individuals and so remains unconfirmed. For example, in group 1 there are seven individuals with very low 813C values relative to 8 15N, and with virtually identical 815N values (Fig. 4B, "sub-group lb"). Of these, five are male/probably male, one is probably female, and one is of indeterminate sex (Roksandic 1999). The female is an elderly individual (>40). Given the age/sex composition of this "sub-group" and the fact that there is at least one child with a similar 813C isotopic signature (see below), it would be difficult to see these as "outsiders" who had married into the Lepenski Vir community (cf. Bonsall et al. 1997). A more likely expla- Similarity 52.23 - 68.15 - 84.08 - 100.00 I ti n h m Vlasac Cluster analysis of the enlarged data set of 35 individuals from Vlasac suggests a division into four groups (Fig. 5A). Groups 1 and 4 are separated on the basis of 515N. Groups 2 and 3 are distinguished on the basis of both 8I3C and S15N (Fig. 5B). 1 2 6 3 4 5 7 8 9 10 1517 11 12 1316)4 18 19 202122 2324 25 262728 2931 32 3330 Observations GROUPS Fig. 3■ Dendrogram grouping 33 skeletons from Lepenski Vir (d. Bonsall et al. 1997^ according to 5l3C and 8I5N values. nation is that they represent a specific phase in the occupation of the site when people had access to aquatic food sources that were relatively depleted in 13C (compared to other group 1 individuals). Variations in the isotopic composition of Danube fish may have occurred through time as a result of natural changes in the freshwater ecosystem. A similar explanation may be proposed for four individuals in group 3 who have unusually enriched 813C relative to 515N values (Fig. 4B, "sub-group 3b"). They evidently consumed larger amounts of S13C-enriched terrestrial (and possibly aquatic) foods compared to other group 3 individuals, which may indicate that they belong to a different phase in the occupation of the site. Theoretically, changes in the natural environment and/or economic practices could have raised average 813C levels of some important food sources. Such changes include (i) an increase in grazing herbivores (including domesticated cattle) at the expense of browsers (deer), as agriculture expanded and woodland cover was reduced; (ii) the introduction of C4 millet (Panicum miliaceum) into the food chain during the Neolithic (or a subsequent increase in its use) either directly as human food or indirectly as grown forage for livestock, and (iii) the appearance of true marine fish (anadro-mous) in the Danube after 6700 BP when the Black Lake was converted into the Black Sea. The suggestion that within their respective groups, "sub-groups" IB and 3B are chronologically distinct is a working hypothesis that requires confirmation from radiocarbon dating. Again, the provisional groups (clusters) do not correspond to archaeological phase (cf. Srejovic and Letica 1978) and groups contain individuals of diverse age at death and both sexes. Of the five radiocarbon ages currently available, four lie in group 1 (9800+108 - 7768+113 BP) and one lies in group 2 (7598±113 BP). The single group 2 age measurement is in trend the youngest and raises the possibility that the two clusters represent different periods in the use of the site. However, to confirm any time relationship between the two groups would require further 14C measurements with improved precision. No 14C age measurements are currently available for groups 3 or 4. Groups 1-3 at Vlasac have 813C and S15N values similar to group 1 at Lepenski Vir, while Vlasac group 4 has §15N values in the range of Lepenski Vir group 2. Vlasac and Lepenski Vir occupy almost identical riverside locations just a few kilometres apart that, presumably, gave access to essentially the same food resources. Therefore, it may be suggested that Vlasac group 4 belongs to the same time-range as Lepenski Vir group 2, and is later than Vlasac groups 1-3 and Lepenski Vir group 1. Children Isotopic analyses are available for 10 children (under 15 years old) from Lepenski Vir. These are compared against the adult ranges in Figure 6. The overall distribution is similar to that of the adults. Six children have 815N values in the range of the group 1 adults, including one with a 813C relative to 815N value reminiscent of the adult "lb sub-group". One child has a 815N value in the range of the group 2 adults, and there are three children whose 815N values are similar to group 3 adults. It is interesting that the first group of children have 815N values that are, on average, l.l%o higher than the corresponding group of adults (+l6.5%o versus +15.4%o). A tendency toward more positive 5I5N values in children compared to adults has been observed in some previous studies (e.g. Pate 1997; Ogrinc 1999). This is usually attributed to the fact that, during infancy, children ingest their mothers' milk and thus, in effect, feed at a higher trophic level. After weaning, it is supposed that bone collagen turnover would result in the progressive loss of this "nursing signal" (cf. Pate 1997). A "nursing effect" is not apparent for the other Lepenski Vir children, perhaps because of the very small number of individuals involved. Nor is it evident among the children from Vlasac (Fig. 7), although the possibility cannot be excluded. Of the seven Vlasac children analysed, five have 813C and 515N values similar to group 1 adults, while another has a 515N value that falls at the bottom of the range for group 2 adults. Assuming a 1.1%0 difference between the average 515N values of adults and children, it is possible that some of the Vlasac children are the offspring of group 4 females, and others are the offspring of group 1 and 2 females. Similarity 43.61 62.41 - 81.20- 100.00 Observations GROUPS B. 18- 16- 1o 14- 12- 10- • Group 1 o Group 2 » Group 3 • • 1b ... 0° 0 o o° 0° A . '' A; A A /A A 3b -22 -21 -20 -19 -18 -17 -16 The remaining child from Vlasac has 515N and 813C values of+12.7%o and -19.8%o, respectively. These are the lowest values recorded for any individual (child or adult) from Vlasac, and fall within the range of the group 2 individuals from Lepenski Vir. This evidence appears to confirm the presence at Vlasac of individuals with "intermediate" diets, and it is not inconceivable that this child was that of a female who had a predominantly terrestrial diet (cf. Lepenski Vir group 3). Because of the possibility of systematic differences between the isotopic signatures of adults and children, it was decided that separate statistical analyses be carried out. Dietary change and the timing of the Mesolithic-Neolithic transition in the Iron Gates The direct AMS 14C age measurements on human bones from Lepenski Vir and Vlasac (Tab. 3) are evi- 5i3c (%o) Fig. 4. A. Dendrogram grouping 58 skeletons from Lepenski Vir according to S'-'C and 5nN values. B. Scatterplot of 5'3C versus 815N for 58 adults from Lepenski Vir, and groupings suggested by cluster analysis. Note the exaggerated scale on the S' s€ axis. dence of human occupation of that part of the Iron Gates gorge from c. 9800-6700 BP. Bone collagen stable isotope analyses indicate fundamental changes in diet during that time range. Humans dated before 7150+106 to 7040+95 BP on the reservoir-corrected time-scale have 815N values of >+14.4%o, indicating diets in which the bulk (>67%) of the protein was derived from Danube fish. This phase is represented by the group 1 individuals at Lepenski Vir and by Vlasac groups 1, 2 and 3. Since there is no evidence for the keeping of domesticated animals (other than dog) prior to that time, it seems reasonable, using this criterion, to describe the pre-7150-7040 BP inhabitants of Lepenski Vir and Vlasac as "Mesolithic". If so, when did the people of the Lepenski Vir-Vlasac section of the Iron Gates become "Neolithic" farmers? A change in diet is evident at Lepenski Vir between 7150-7040 BP. The group 2 individuals from Lepenski Vir show a significant reduction in average 815N values, consistent with an increase in the amount of protein derived from terrestrial food sources and a corresponding decrease in protein from aquatic sources. The same change may be represented at Vlasac by the group 4 adults. At Lepenski Vir this can be seen as the beginning of a trend that culminated in the adoption of a predominantly terrestrial diet by c. 6960 BP, represented by the group 3 individuals. The timing of this dietary change corresponds quite closely with the appearance of Neolithic farmers in the regions surrounding the Iron Gates, represented by the earliest Starcevo-Cris-Koros settlements, and it is reasonable to infer that the two events are connected. There are two hypotheses that can plausibly account for the changes observed at Lepenski Vir (and possibly Vlasac) after 7150+106 BP. The first is that Mesolithic people of the Lepenski Vir-Vlasac area adopted farming more or less as soon as it became available to them, and gradually increased the amount of agricultural products in their diets at the expense of traditional aquatic resources. The second is that the local population did not become farmers immediately, but traded with neighbouring farmers for agricultural products for a period of decades to centuries before eventually taking up livestock raising and cultivation. Similarity 51.60 67.74 83.87 - 100.00 3 9 15 17 5 27201128 6 10 4 1626 1819 1 14121330 2 2531243423 8 2229 7 21333235 Observations i_i i_i i_i i_i 1 GROUPS 17.0 B. 16.5- 16.0- 15.5 J 15.0- tO 14.5- 14.0- 13.5- 13.0- 12.5 x Group 1 ■ Group 2 o Group 3 • Group 4 o o x -20.0 -19.5 -19.0 -18.5 -18.0 813C (%O) Fig. 5- A. Dendrogram grouping 35 skeletons from Vlasac according to ft^C and 8nN values. B. Scatterplot of S*3C versus S'\\for 35 adults from Vlasac, and groupings suggested by cluster analysis. This latter possibility has been suggested by several authors, most notably Voytek and Tringham (1989). On the existing radiocarbon evidence, an "availability phase" (cf. Zvelebil and Rowley-Conwy 1984; 1986) during which Mesolithic people in the Iron Gates increased their intake of terrestrial protein through trade or exchange with farmers, could have lasted between a few decades and approximately 600 years. The same radiocarbon evidence suggests that it would have ended by c. 6800 BP at the latest. Two individuals from Lepenski Vir group 3 with reservoir-corrected radiocarbon ages of 6960+93 BP and 6720+93 BP have S15N values of+10.9%o and +11.2%o, respectively, indicating predominantly (64-68%) terrestrial diets. In the context of the Iron Gates, it is difficult to see how such high levels of terrestrial protein intake could have been sustained without a direct investment in agriculture. 18 \C - 16- 14- - 12 CO 10 • 1b.......... ......•• • * # ha 2........ J ° 3a" .....A •* A (...-•"3b -22 -21 -20 r -19 513C (%O) -18 -17 -16 Fig. 6. 5l3C versus 815N for 10 children from Lepenski Vir, plotted against ranges of adult groupings suggested by cluster analysis. Other explanations could be proposed for the initial reduction in average 8HN values at 7150-7040 BP (cf. Lepenski Vir group 2, Vlasac group 4). They include a long-term increase in the consumption of wild animal and/or plant resources, a reduction in the average size of freshwater fish caught, or even a change in the type of fish caught. However, there is no convincing supporting evidence from either Lepenski Vir or Vlasac. Even if there were, such changes are most unlikely to account for the strongly "terrestrial" isotopic profile of the group 3 individuals from Lepenski Vir. Stable isotopes and the dating of houses at Lepenski Vir Since the publication of the account of the Lepenski Vir excavations (,Srejovic 1969; 1972) there has been considerable controversy over the age and cultural context of the trapezoidal-plan houses of Lepenski Vir I and II. While most authors now accept the radiocarbon ages on charcoal from the houses as valid, there is still disagreement on whether the houses should be interpreted as Mesolithic (cf. Radovanovic 1996) or Neolithic (cf. Ehrich 1974; Milisauskas 1978). The radiocarbon and stable isotope measurements on human bones from Lepenski Vir, discussed in this paper, have a critical bearing on the issue. Charcoal samples from the LV I—II houses produced ]4C ages between 6560 and 7430 BP (Quitta 1972; Boric 1999). These are similar to the reservoir-corre-cted 14C ages on human bone (Tab. 3). If the 14C ages of the charcoal and human bone samples are accepted as being correct, then the houses and the human bones can be considered as belonging to approximately the same time-range. Since the human bones appear to span the transition from a Mesolithic to a Neolithic economy, it would be reasonable to conclude that the same applies to the houses. However, as Cook et al. (in press and forthcoming) have pointed out, the charcoal samples were from long-lived tree species (oak and elm). Such samples can yield l4C ages that are several hundred years older than the archaeological events they purport to date - often referred to as the "old wood" problem. Therefore, it is conceivable that the houses are significantly younger than the radiocarbon ages of the charcoal samples, and all the houses post-date the change in diet between 7150 and 7040 BP, i.e. they belong to the time-range of the group 2 and 3 humans. This would be consistent with the presence of Starcevo pottery in several of the houses that were dated (Budja l999.Fig. 7). These are house 54 (7l6l±56 BP - weighted mean of five 14C measurements), house 1 (6860+100 BP) and house 16 (6820+100 BP). Given the uncertainties over the interpretation of the charcoal-based radiocarbon ages, the question may be asked: can it be shown that any of the houses at Lepenski Vir belongs to the period of the group 1 ("Mesolithic") humans? It was suggested above that the division of the human remains from Lepenski Vir into three groups, according to the 813C and 815N values, represents a time-series. For convenience, this may be characterised as: Period 1 ("Mesolithic") dating before 7040± 95 BP and comprising individuals with 815N values of —+l4.4%o. Period 2 ("transitional Mesolithic-Neo-lithic") dating c. 7040+95 BP and represented by individuals with SHN values ranging between +11.8 to +I4.0%o. Period 3 ("Neolithic") dating after 7040± 95 BP and comprising individuals with S15N values of <+11.2%o. If the "phasing" of the human remains based on 515N is reliable, where there is a clear stratigraphic relationship between a human skeleton and a house, it follows that the bone collagen 515N value can be used to infer the age of the house. However, this would only apply in the case of articulated skeletons. Many of the "skeletons" uncovered in the Lepenski Vir excavations appear to be groups of disarticulated bones. These could represent delayed or secondary burials of individuals who had died some time previously, and hence there could be a significant "age offset" between the time of death of the individual and the time of final burial. Few details of the stratigraphic relationships of burials and houses are provided in published accounts of the Lepenski Vir, but there is a limited amount of photographic evidence that can be considered. Published photographs of house 21 (Srejovic 1969. PL 69; Radovanovic 1996.Fig. 4.3) show the articulated skeleton of an adult female lying below the floor of the house. It is clear from other photographs published by Radovanovic (1996.Figs 3-14, 3-31, 3-32) that the burial was inserted through the plaster floor. This relationship indicates that the burial must have been emplaced after the plaster floor was laid, and therefore (presumably) post-dates construction of the house. The skeleton (7b or 7/1) has a 815N value of +15.8%o, placing it firmly within Period 1 ("Mesolithic"). Unless this is a case of delayed burial following excarnation, which seems highly improbable, the evidence implies that house 21 is also Mesolithic. According to Srejovic {1969; 1972; Srejovic and Ba-bovic 1983) and Radovanovic (1996) house 21 is superimposed upon houses 22, 29 and 30 (Fig. 8), which therefore places them also in period 1. Thus, there are at least four houses at Lepenski Vir that, on the combined evidence of stratigraphy and bone collagen isotopic data, can be argued to pre-date the change in diet between 7150 and 7040 BP. 17.0- 16.5- 16.0- 15.5- ^_s o a? 15.0- z VI 14.5- CO 14.0- 13.5- 13.0- 12.5- 3 1 f • • • 4 '''■•■... • -20.0 -19.5 -19.0 Sl3C (%o) -18.5 Fig. 7. 5I3C versus 8'5N for 7 children from Vlasac, plotted against ranges of adult groupings suggested by cluster analysis. It is significant that none of the four probable period 1 houses discussed appears to have been associated with Starcevo pottery (cf. Budja 1999.Fig. 7). This would be consistent with a Mesolithic context and an age prior to 7040 BP. The photograph in Srejovic (1969.Pl. 67; Srejovic 1972.PI, 58) shows the articulated skeleton of an adult female (54e) lying directly above stone slabs apparently set into the floor of house 65 (Fig. 8). This skeleton has a 815N value of +13.2%o and belongs to period 2, post-dating the dietary change at 7150-7040 BP. Leaving aside the question of whether the corpse was deliberately left exposed on the floor of the house, or placed beneath a cairn (a possibility suggested by other photographic evidence), or buried in a grave pit dug from a higher level, the position of the skeleton with respect to the floor suggests that the house is older and could belong to either period 1 or period 2. A period 2 (post-7150-7040 BP) age would be consistent with the presence of pottery inside the house (cf. Budja 1999-Fig. 7). It was suggested by Srejovic that house 65 contained an earlier burial (54d), which is represented by disarticulated bones around skeleton 54e (Srejovic 1969.Pl. 67; Srejovic 1972.Pl. 58). One of these bones gave a 815N value of +15.3%o suggesting an age prior to 7150-7040 BP. It was claimed that the bones of 54d had been disturbed by burial 54e. However, given the disarticulated nature of 54d and the fact that the bones may be from more than one individual, they are more plausibly interpreted as a secondary burial and cannot therefore be used to "date" house 65. The stable isotope data also have implications for Radovanovic's architectural phasing of the Lepenski Vir houses (Radovanovic 1996). As noted above, there is good evidence that houses 21, 22, 29 and 30 are Mesolithic and predate 7150-7040 BP. In Radovanovic's scheme, houses 21 and 22 are assigned to phase 1.2, and houses 29 and 30 to phase 1.1 (Fig. 8). If her phasing of the houses were correct, one would have to conclude that the shift away from a traditional Mesolithic diet began either during or after phase 1.2, and that all houses assigned to phase 1.1 are Mesolithic. However, this interpretation is inconsistent with radiocarbon evidence from three other houses attributed to phase 1.1 (houses 1, 9 and 37) suggesting -18.0 65601100 BP 66 (phase 1.3) under 66201100 BP XXXII, 6859164 BP 7430+160 BP 6820+100 BP 6860+100 BP 7335171 BP 6845+100 BP 6970+60 BP uncertain Danube ^,66301100 BP IX ^ 72101200 BP Fig. 8. Lepenski Vir I site plan showing architectural phases and radiocarbon ages of houses based on charcoal samples (after Radovanovic 1996, with acknowledgement to Boric 1999 and Bailey 2000/ The UC ages of houses 36, 37 and 54 are weighted means of several measurements. The locations of two Lepenski Vir II houses (IX and XXXII) and corresponding 14C ages are also shown. that they post-date the dietary change at 7150-7040 BP (Fig. 8). The use of stable isotope data as a proxy dating tool may also contribute to a better understanding of the evidence from Vlasac. Five charcoal samples from Srejovic and Letica's (1978) phase lb gave 14C ages of 6805-7000 BP. These ages were rejected because they were out of sequence with radiocarbon determinations for the succeeding phases II and III, and because they were not in accord with the excavators' belief that the contexts dated were Mesolithic. However, in European archaeology radiocarbon ages have often proved more reliable than chronologies derived from archaeological observations. The ages for "Vlasac lb" are consistent with the presence of Early Neolithic (Starcevo) remains on the site. These have always been considered a very minor component of the archaeological record. However, as noted above, stable isotope evidence indicates that a significant proportion of the humans buried at Vlasac - the group 4 adults, representing 23% of the samples analyzed - had diets similar to the group 2 adults from Lepenski Vir, and may therefore belong to the same time-range of c. 7040±95 BP. CONCLUSIONS Reappraisal of a larger stable isotope data set for Lepenski Vir and Vlasac demonstrates a shift from a Mesolithic-type dietary regime, based largely on aquatic resources, through an intermediate phase, to one based largely on terrestrial resources that probably included a major agricultural component. Radiocarbon evidence suggests that the transition centred around 7040±95 BP (6156-5721 cal BC), and that agriculture was being practised in the Lepenski Vir-Vlasac area by 6800 BP (c. 5700 cal BC). The stable isotope and radiocarbon data coupled with evidence of the stratigraphic relationships between burials and houses suggest that the trape- zoidal plan houses of "LV I—II" span the time-range of the dietary change. This is contrary to previous interpretations of the houses as either exclusively Mesolithic or exclusively Neolithic. Moreover, the stable isotope evidence suggests that Lepenski Vir (and possibly Vlasac) was occupied continuously from the Mesolithic into the Early Neolithic. For both Lepenski Vir and Vlasac, this paper has also highlighted apparent conflicts between archaeological sequences and radiocarbon based chronologies supported by stable isotope analyses. Future research must be directed toward resolving the issues that have been raised here. REFERENCES AMBROSE S. H. 1993- Isotopic analysis of paleodiets: methodological and interpretive considerations. In Sandford M. K. (ed.), Investigations of Ancient Human Tissue: Chemical Analyses in Anthropology: 59-130. BAILEY D. W. 2000. Balkan Prehistory. London: Routledge. BONSALL C., LENNON R., McSWEENEY K., STEWART C., HARKNESS D., BORONEANT V., BARTOSIEWICZ L, PAYTON R. and CHAPMAN J. 1997. Mesolithic and Early Neolithic in the Iron Gates. A palaeodietary perspective, fournal of European Archaeology 5: 50-92. BORIC D. 1999- Places that created time in the Danube Gorges and beyond, c. 9000-5500 BC. Docu-menta Praehistorica XXVI, 6"' Neolithic Studies: 41-70. BOKONYI S. 1969. A Lepenski Vir-i oskori telep ge-rinces faunaja (The vertebrate fauna of the prehistoric settlement at Lepenski Vir). Archeologiai Erte-sfto 96:157-160. 1970. Animal remains from Lepenski Vir. Science 167, 3926: 1702-1704. 1977. Vlasac und die Fragen der mesolithischen Domestikation. Mitteilungen des Archaologischen Instituts der Ungarischen Akademie der Wissen-schaften 7: 85-92. 1992. Animal remains of Mihajlovac-Knjepiste, an Early Neolithic settlement of the Iron Gate Gorge. (Hommage a Nikola Tasic a l'occasion de ses soix-ante ans), Balcanica XXIII: 77-87. BUDJA M. 1999. The transition to farming in Mediterranean Europe - an indigenous response. Docu- menta Praehistorica XXVI, 6th Neolithic Studies: 119-141. COOK G. T, BONSALL C., HEDGES R. E. M., McSWEENEY K., BORONEANT V., PETTITT P. B. In press. A freshwater diet-derived 14C reservoir effect at the Stone Age sites in the Iron Gates gorge. Radiocarbon 43: accepted for publication. COOK G. T, BONSALL C., HEDGES R. E. M., McSWEENEY K, BORONEANT V, BARTOSIEWICZ L. and PETTITT P. B. Forthcoming. Problems of dating human bones from the Iron Gates. Submitted to Antiquity. DUFOUR E„ BOCHERENS H. and MARIOTTI A. 1999. Palaeodietary implications of isotopic variability in Eurasian lacustrine fish Journal of Archaeological Science 26: 617-627. EHRICH R. W. 1974. Book review of Hunters, Fishers and Farmers of Eastern Europe 6000-3000 BC, by Ruth Tringham. American fournal of Archaeology 78: 435-436. HEDGES R. E. M., PETTITT P. B, BRONK RAMSEY C. and VAN KLINKEN G. J. 1998. Comment (Lab.), p. 440. In Radiocarbon dates from the Oxford AMS System: Archaeometry datelist 26. Archaeometry 40(2): 437-455. HOBSON K. A. and WELCH H. E. 1995. Cannibalism and trophic structure in a high Arctic lake: insights from stable-isotope analysis. Canadian fournal of Fisheries and Aquatic Science 52: 1195-1201. IACUMIN P., BOCHERENS H., CHAIX L. and MARIOTH A. 1998. Stable carbon and nitrogen isotopes as dietary indicators of ancient Nubian populations (Northern Sudan), fournal of Archaeological Science 25: 293-301. ' KLEPPINGER L. L. 1984. Nutritional assessment from bone. Annual Review of Anthropology 13: 79-96. MAYS S. 1998. The Archaeology of Human Bones. London: Routledge. MILISAUSKAS S. 1978. European Prehistory. New York: Academic Press. MINAGWA M. and WADA E. 1984. Stepwise enrichment of 15N along food chains: further evidence and the relation between 815N and animal age. Geochi-mica et Cosmochimica Acta 48:1135-1140. OGRINC N. 1999. Stable isotope evidence of the diet of the Neolithic population of Slovenia - a case study: Adjovska jama. Documenta Praehistorica XXVI, 6th Neolithic Studies: 193-200. PATE F. D. 1997. Bone chemistry and paleodiet: reconstructing prehistoric subsistence-settlement systems in Australia, fournal of Anthropological Archaeology 16: 103-120. 1998. Bone collagen stable nitrogen and carbon isotopes as indicators of past human diet and landscape use in southeastern South Australia. Australian Archaeology 46: 23-29. QUITTA H. 1972. The dating of radiocarbon samples. In D. Srejovic (ed.), Europe's First Monumental Sculpture. New Discoveries at Lepenski Vir: 205-210. RADOVANOVIC I. 1996. The Iron Gates Mesolithic. Ann Arbor: International Monographs in Prehistory, Archaeological Series 11. ROKSANDIC M. 1999. Transition from Mesolithic to Neolithic in the Iron Gates Gorge: Physical Anthropology Perspective. PhD Thesis, Simon Fraser University. RYAN W. B. F, PITMAN W. C., MAJOR C. 0, KAZIMIE-RAS S„ MOSKALENKO V., JONES G. A., DIMITROV P., GORUR N., SAKINg M. and YUCE H. 1997. An abrupt drowning of the Black Sea shelf. Marine Geology 138: 119-126. SCHULTING R. J. 1999- Slighting the sea: stable isotope evidence for the transition to farming in northwestern Europe. Documenta Praehistorica XXV, 5th Neolithic Studies: 203-218. SCHWARCZ H. P. 1991- Some theoretical aspects of paleodiet studies Journal of Archaeological Science 18: 261-275. SREJOVIC D. 1969. Lepenski Vir - Novapraistorij-ska kultura u Podunavlju. Beograd: Srpska knjizev-na zadruga. 1972. Europe's First Monumental Sculpture. Neiv Discoveries at Lepenski Vir. London: Thames and Hudson. SREJOVIC D. and BABOVIC L. 1983- Umetnost Le-penskog Vira. Belgrade: Yugoslavija. SREJOVIC D. and LETICA Z. 1978. Vlasac. Mezolit-sko naselje u Djerdapu (Vlasac. A Mesolithic Settlement in the Iron Gates). Beograd: Serbian Academy of Sciences and Arts Monograph DXII. STUIVER M. and REIMER P. J. 1993. Extended 14C database and revised CALIB radiocarbon calibration program. Radiocarbon 35: 215-230. STUIVER M, REIMER P. J., BARD E., BECK J. W, BURR G. S., HUGHEN K. A., KROMER B, McCORMAC F. G, VAN DER PLICHT J. and SPURK M. 1998. INTCAL98 radiocarbon age calibration, 24,000-0 cal. BP. Radiocarbon 40: 1041-1083- VOYTEK B. A. and TRINGHAM R. 1989- Rethinking the Mesolithic: the case of southeast Europe. In C. Bonsall (ed.), The Mesolithic in Europe, 492-499. ZOFFMANN Z. K. 1983- Prehistorical skeletal remains from Lepenski Vir (Iron Gate, Yugoslavia). Homo 34: 129-148. ZVELEBIL M. and ROWLEY-CONWY P. 1984. Transition to farming in northern Europe: a hunter-gath-erer perspective. Norwegian Archaeological Review 17: 104-128. 1986. Foragers and farmers in Atlantic Europe. In M. Zvelebil (ed.), Hunters in Transition: 67-93■