The Emergence of Dairying in Early Farming Practices of the Fertile Crescent and the Balkans


Principal investigators: Professor Richard P. Evershed, Professor Andrew G. Sherratt & Dr Sebastian Payne ; Postdoctoral researchers: Drs Mark S. Copley & Jen Coolidge



A Leverhulme Trust funded project is aiming to determine the timing of the emergence of dairying in early farming practices through analyses of milk fat residues in pottery using a compound-specific stable isotope method developed by the Bristol group ( Science 1998) and previously applied to the study of the importance of dairying in prehistoric Britain ( Proceeding of the National Academy of Sciences USA 2003). This new project will be the largest regional study ever undertaken using organic residues to answer an archaeological question. Since we will be drawing on pottery from sites across a wide geographic and temporal range, we have sought the cooperation of a large number of archaeologists, which includes the site of Çatalhöyük.


Leverhulme Projesi tarafından fonu sağlanan proje, Bristol grubu ( Science 1998) tarafından bulunan ve daha önce prehistorik Britanya'da süt ve sütçülük ile ilgili çalışmada uygulanan ( Proceeding of the National Academy of Sciences USA 2003) bileşik-spesifik stable izotop yöntemi ile çanak-çömleklerin dibindeki süt kalıntılarını inceleyip erken tarım topluluklarında süt kullanımının ortaya çıkışını araştırmaktadır. Bu yeni proje, simdiye kadar bir arkeolojik soruyu yanıtlamak adına organik kalıntı kullanılan, bölgesel anlamdaki en geniş projedir. Çok geniş bir alana yayılmış çanak-çömlek kalıntılarının incelenmesi yapılacağından, içlerinde Çatalhöyük arkeologlarının da bulunduğu geniş bir arkeolog topluluğu ile kooperasyon kurmuş bulunuyoruz.



Dairying in prehistory
The domestication of the principal livestock species (goat, sheep and cattle) took place in the Near (Middle) East in the 8th millennium BC, in the area known as the "Fertile Crescent". By the 4th millennium BC there is evidence for the specialised exploitation of these animals for milk, wool and traction purposes by the first urban communities in Mesopotamia (Sherratt, 1981) , principally indicated by textual and pictorial evidence. The initial (Neolithic) mode of exploitation in this area, as shown by culling patterns reconstructed from bones recovered in excavations, suggests an emphasis on young animals for meat, and this simple type of livestock use is typical of the earliest farming groups in Europe (Bogucki, 1984, Benecke, 1994). However, there is considerable uncertainty about the date at which milking was first practised, and this is particularly interesting in view of the widespread occurrence of adult lactase deficiency in human populations. By the time that farming reached the British Isles in the 4th millennium BC, herd structures with a high proportion of mature females indicate that cattle were probably being exploited for dairy products (Legge, 1981) . This supports a reconstruction of the beginnings of dairying as shown below (Fig. 105).


Figure 105: Model for the emergence and spread of farming and dairying across Europe in the Neolithic. The milking of animals in the Fertile Crescent is hypothesised to have occurred some considerable time after their initial domestication. By the time farming reached NW Europe, dairy products were an important component of farming practices (as we demonstrate in Figure 107).


It has recently been shown to be possible to provide direct chemical evidence for the processing of milk products by stable isotope analysis of organic residues preserved in pottery vessels from all prehistoric periods in Britain as far back as the Neolithic (Dudd and Evershed, 1998 ; Copley et al., in prep). Given the success of this approach and its general applicability to the analysis of dairy fats in potsherds from any region the possibility now exists to extend this investigation to earlier farming groups in South-east Europe and the Near East and test the hypothesis regarding the timing of the beginnings of dairying. This would resolve a major question in human dietary history and the development of animal exploitation in Europe and the Near East.


Chemical analysis of organic residues in pottery

During the processing of organic material in unglazed pottery vessels, e.g. the cooking of food, lipids are absorbed into the vessel wall (e.g. Evershed et al., 1999) . These lipids have been shown to survive over considerable archaeological time periods, and can be chemically extracted and characterised using modern analytical chemical techniques including gas chromatography (GC) and GC/mass spectrometry (GC/MS). We have used these techniques to provide quantitative estimates of the amounts of lipid present. More importantly, they also provide chemical ‘fingerprints' based on the structures of individual compounds and their compositions (or ‘distributions') which have enabled us to identity a variety of substances derived from commodities processed in the vessels, including plant leaf waxes and degraded animal fats (e.g. Evershed et al., 1991, Evershed et al., 1999) . Furthermore, we have also been able to define the origins of degraded animal fats via the stable carbon isotope ( d 13 C) values determined by GC-combustion-isotope ratio mass spectrometry (GC-C-IRMS) of the individual fatty acid components (hexadecanoic and octadecanoic fatty acids; C 16:0 and C 18:0 , respectively). Through this approach non-ruminant adipose fats, ruminant adipose fats and ruminant dairy fats have been distinguished (Evershed et al., 1997, Dudd and Evershed, 1998) .


Molecular and stable isotope evidence for dairy fat residues in pottery

As is the case with any other fat/oil, the processing of milk, e.g. pasteurising, or cooking involving milk or butter, would result in the absorption of significant quantities of lipid into the walls of unglazed pottery vessels. Milk fats differ from adipose fats in their fatty acid composition through the presence of short chain saturated fatty acyl moieties in the C 4 to C 14 carbon number range (McDonald et al., 1988) . Degradation experiments in our laboratory have demonstrated that these short fatty acyl moieties are more susceptible to hydrolytic degradation due to reduced steric effects at ester linkages in triacylglycerols compared with their longer chain counterparts (Dudd and Evershed, 1998) . Furthermore, once released from triacylglycerols by hydrolysis, the short-chain fatty acids are appreciably more water-soluble than their long-chain counterparts; significantly there is a c. 10 fold decrease in solubility for each added methylene group (Bell, 1973) . These two factors alone are sufficient to explain why short-chain fatty acids do not survive in ancient pottery vessels.

However, the C 16:0 and C 18:0 fatty acids are commonly observed in appreciable abundance in archaeological ceramics, and it is these components that are particularly useful in the differentiation of ruminant and non-ruminant adipose fats, and in the identification of dairy fats. Distinguishing between ruminant milk and adipose fats rests on the higher d 13 C values of C 18:0 in dairy fats as shown in Figure 106a, which results from: (i) the lack of capacity of the mammary gland to biosynthesise C 18:0 (Christie, 1981, Byers and Schelling, 1988) , (ii) the different isotope values of lipids and carbohydrates in plants (Fig. 106b) resulting from fractionation during the decarboxylation of pyruvate in forming acetylCoA (De Niro and Epstein, 1977) , and (iii) the routing of c. 40% of dietary C 18:2 and C 18:3 fatty acids after biohydrogenation to C 18:0 in the rumen, to milk during lactation (McDonald et al., 1988) . This results in the d 13 C value of the C 18:0 fatty acid in ruminant dairy fats being c. 2.3‰ more depleted than in ruminant adipose fats (Fig. 106c.), and this occurs irrespective of whether the animal's diet comprised of C 3 , C 4 or mixtures of C 3 and C 4 plants. Recognition of this latter phenomenon and extensive research into the fats of reference animals and their diets, underpin our current research aimed at exploring the importance of dairying in prehistory. Equally importantly, the d 13 C values of the C 16:0 and C 18:0 have been shown, through long-term degradation experiments, to be robust criteria that are unaffected by degradation during burial of potsherds.


Figure 106. (a) Plot of the d 13 C values of the major saturated fatty acid components (C 16:0 and C 18:0 ) of the lipid extracts from modern reference fats showing the resolution of dairy fats from ruminant and non-ruminant adipose fats. The three fields correspond to 1 s.d. confidence ellipses calculated for the d 13 C values of the domesticates known to comprise the major component of prehistoric economies in Britain. All the animals were raised on C 3 diets. The d 13 C values for the fatty acids have been corrected for post-Industrial Revolution effects of fossil fuel burning, which has led to more depleted atmospheric d 13 CO 2 values, and has been calculated to be 1.2‰ (Friedli et al., 1986) . Analytical precision is ±0.3‰. (b) Histogram of the d 13 C values of C 18:3 fatty acids and glucose extracted from plants. The histogram of the d 13 C values of the major fatty acids and carbohydrates of 166 modern plants demonstrates that there is an 8.1‰ mean difference in the d 13 C values of C 18:3 fatty acid (mean = -36.3‰) and glucose (mean = -28.2‰) which is the basis of the difference in the d 13 C value of the C 18:0 fatty acid in dairy and milk fat. This difference in d 13 C value between lipids and carbohydrates is seen in both C 3 and C 4 plants. (c) Diagram showing the routing of dietary fatty acids and carbohydrates in the rumen, adipose tissue and mammary gland of the ruminant animal. 60% of the C 18:0 in ruminant milk is rerouted from adipose tissue and is comprised of carbon originating from dietary glucose and fatty acids, and the remaining 40% (McDonald et al., 1988) is directly incorporated from the diet following biohydrogenation of unsaturated fatty acids in the rumen (* in diagram), and reflects the inability of the mammary gland to biosynthesise C 18:0 (Christie, 1981, Byers and Schelling, 1988) . The 2.3‰ mean difference in the d 13 C values of C 18:0 in ruminant adipose tissues and dairy fats can be seen graphically in Fig. 106a.


Evidence for dairying in Neolithic Britain

We have already determined the extent of dairying in prehistoric Southern Britain using these new molecular and stable isotopic methods; over 800 sherds were analysed from twelve sites ranging from the early Neolithic to late Iron Age (Copley et al. in prep). Approximately 50% of these sherds yielded lipids, allowing the d 13 C values of the fatty acids to be determined. Figure 107 shows the results of organic residue analysis from three specific Neolithic sites in Britain. The faunal evidence at both Windmill Hill and Hambledon Hill is suggestive of dairying (Grigson, 1999 , Legge, in prep) whereas the faunal assemblage at Eton Rowing Lake is more fragmentary. However, dairy fats were observed in significant numbers of sherds from all three sites. This illustrates the importance of residue analysis: (i) where the faunal assemblage is relatively fragmentary, (ii) as a means of directly demonstrating the extent of processing of dairy fats in pottery, and (iii) in the determination of the relative frequency of ruminant/non-ruminant animal fats in the sherds. The results from all twelve British prehistoric sites are summarised in Table 1/56. These data clearly demonstrate that dairying was an important component of prehistoric economies from the early British Neolithic onwards (c. 4000BC -), and that the milking of animals was introduced in Britain at the same time as farming itself.


Figure 107:. Plot of the d 13 C values of the major fatty acid components (C 16:0 and C 18:0 ) of lipid extracts from potsherds from the Neolithic sites of (a) Windmill Hill, (b) Hambledon Hill and (c) Eton Rowing Lake. The fields corresponding to the modern reference fats (c.f. Fig. 104 a) serve to classify the lipid extracts. Sherds that plot between the ruminant and non-ruminant adipose fat ellipses provide evidence for the mixing of these fats in the vessels. Analytical precision is ±0.3‰.


Table 56: Summary of site from which pottery was analysed showing importance of dairying in prehistoric Britain


Organic residues in Neolithic Near Eastern pottery
In order to assess the level of preservation of lipids in very early pottery vessels from the Near East, pilot studies that have been completed, utilising sherds from Çatalhöyük, Turkey and from Tell Bouqras and Tell Sabi Abyad, Syria. Results obtained from sherds from these sites have shown that degraded animal fats are preserved in pottery vessels dated to the 6 th Millennium BC. Out of the 40 sherds from these three sites, 10 (25%) have yielded significant concentrations (>50 m g g -1 ) of lipid. Fig.108 shows two typical gas chromatograms obtained during these pilot studies. These three sites are over 2000 years older than the early Neolithic pottery we have analysed from Britain. This is probably oldest pottery in which lipid residues have been reliably detected and given the great age of the vessels the level of preservation is remarkable. These data confirm that results will be obtained from other pottery throughout the regions targeted in this proposal.



Figure 108. Partial high temperature gas chromatogram of the total lipid extract from a sherd from Çatalhöyük, Turkey (top) and Tell Bouqras, Syria (bottom). C X:Y refer to fatty acids of carbon chain length x and level of unsaturation y. C x MAGs are monoacylglycerols of acyl carbon number x; C x DAGs are diacylglycerols of acyl carbon number x; C x TAGs are triacylglycerols all of carbon chain length x. IS is the internal standard (n-tetratriacontane). P is a plasticiser originating from the plastic bag that the sherd was kept in during storage. A DB1 (immobilised dimethyl polysiloxane) fused silica capillary column (15 m x 0.32 mm, 0.1 m m film thickness) was used. The chromatograms are dominated by saturated fatty acids, which is in keeping with organic residue analysis of archaeological pottery from other geographical regions that we have studied. However, varying abundances of TAGs, DAGs and MAGs are present, thus indicating the remarkable level of preservation that exists at these sites, even after 8000 years of burial.

Hypotheses to be tested

The principal question is to determine the date at which milk and milk-products came to be a regular part of human diet. It is currently hypothesised (see Fig. 105), on the basis of indications from the reconstructed herd-structure of domestic livestock populations, that the use of milk began one or two millennia after the beginnings of farming and domestication of livestock. This programme will be the first direct test of that reconstruction. In addition, samples have been selected to test whether the use of milk products first began where farming was pioneered, within the Fertile Crescent, or whether it was an innovation of temperate regions (central and south-east Europe).



It is proposed to investigate pottery from a number of systematically selected archaeological sites from the Near East (including the Fertile Crescent) dating from 7500 BC, and from the adjacent part of south-east (Balkan) Europe dating from 6000 BC. A range of sherds from each site will be submitted to organic residue analysis, with the possibility of supplementary analyses to widen the sample or to investigate promising patterns. This programme of research will throw light on the following aspects of early farming economy and diet:

The beginning of the use of domesticated animals for milk and milk-products. It has been argued (Sherratt, 1981, Sherratt, 1983) that the keeping of livestock for secondary products represents an advanced phase of animal domestication within the Fertile Crescent. Investigation of pottery from well-stratified early contexts (mid 8 th to early 5 th Millennium BC), in conjunction with specialist study of the faunal evidence from sites like Çatalhöyük, will be the first objective test of this idea.

The importance of dairy products in the regional economy. The spread of samples across Neolithic farming economies in western Asia and the Balkans may be expected to reveal regional differences in livestock use and diet, and in the importance of milk products in the economy.

The dating and geographical context of early uses of milk. The extent to which milk and milk-products were used by early populations has important implications for questions such as the evolution of lactose tolerance.

This project will thus apply recently developed sophisticated molecular and stable isotope methods of organic residue analysis of proven value, and by extending them to critical areas in the development of farming promises to provide objective answers to major questions in the history of human ecology and diet.



Although at some sites the faunal assemblage may suggest a herd structure associated with dairying (i.e. high neonatal cull and high abundance of mature females), due to inevitable biases against the survival of small bones in the burial environment, as well as particular fragmented assemblages in acid soils, more direct methods for the detection of dairy products are needed. Following on from the successful research of British Neolithic pottery, this work will provide the first direct chemical evidence for the processing (or otherwise) of dairy products in S.E. Europe and The Near East. Therefore questions regarding the timing of the first exploitation of animals for dairy products can finally be answered.



The research programme will involve the analysis of at least 600 potsherds from at least eight archaeological sites (divided equally between The Near East and S.E. Europe) together with parallel analyses of reference plant materials and animal fats.

Site and pottery selection
The major considerations in the selection of pottery for this investigation are:

The availability of sherds in sufficient numbers; in practice, this places temporal limits of post-6,500/6,000 BC, as earlier periods are essentially aceramic.

In order to accurately determine the beginning of dairying we will select sherds from well-stratified and well-dated contexts.  

The need to sample large numbers of sherds from each assemblage, i.e. >60, places a practical limit of 8-12 assemblages.

The selection of the sites from the different regions are based on the following specific criteria:

Those possessing good faunal/botanical assemblages, with specific emphasis on sites where published bulk isotope values of the bones exists.

Those displaying special preservation conditions, such as waterlogging or desiccation.

Sites for which exceptional quantities of relevant material exist, thus allowing particular vessel types to be targeted individually to demonstrate their particular use.

Evidence of widespread cultural groups with characteristic ceramic assemblages well described in the literature.

The applicants already have access to the appropriate excavated pottery samples from known contexts, the majority of which are already in the UK, and extensive use will be made of relevant material in British museums (for instance very large pottery assemblages are available in Oxford from 6th millennium sites in Hungary and 5th millennium sites in Bulgaria and Romania). Opportunities also exist for pottery to be sampled from new excavations during the course of the project. Table 2/57 indicates the chronological range of sites from which samples are already available, either supplied by the excavators or available in UK museum collections for which sampling permission has already been given. If preservation levels in the selected assemblages fall below the expected ranges (which highly unlikely), it will be possible to supplement them with samples from deposits in the circum-Alpine region, from museums in Zurich, Ljubljana and elsewhere.

The sampling will be designed to reflect the range of contemporary vessel forms and their different roles in food-processing and consumption practices. For this purpose it will be necessary to create a set of explicit categories of vessel-types, to which sampled vessels and sherds can be assigned. These functional categories need to be defined from vessel capacity (which can be estimated directly or from rotation of the interior profiles), handle or lug types, aperture size (eg rim-diameter or equivalent), wall thickness and variability, fabric-type (eg temper type and thermal characteristics), together with surface-finish and use-traces (eg sooting). The sampled assemblages need to be compared in these respects with contemporary ranges of variability reported in the literature, and may be summarised visually in diagrams such as that reproduced in Sherratt (1997, 15.6), to highlight contrasts between the container-assemblages characteristic of successive periods.


Table 57: Summary of sites from which pottery will be sampled, showing the time span covered


Molecular and isotopic analyses of absorbed organic residues from potsherds

At least 600 sherds will be selected, from at least four sites from the two regions. The sampling strategy employed during our previous analyses of European pottery has involved the selection of vessels with forms consistent with what may be termed ‘cooking pots', i.e. they are most likely to have been used in the processing of foodstuffs. The success of this strategy for the detection of dairy fats (see Fig. 107; Table 1/56) is due to the fact that the processing of milk, e.g. pasteurising, or cooking involving milk or butter, would result in the absorption of significant quantities of lipid into the walls of unglazed pottery vessels. Specialist vessels, such as ceramic sieves and butter churns do not form the basis of this application, but examples will be considered where they are available.

The lipid components of small samples (ca. 2 g) of sherds will be analysed according to our published procedures (Evershed et al., 1997; Dudd and Evershed, 1998; Mottram et al., 1999). Briefly, after solvent extraction of powdered sherds, lipid extracts will be converted to trimethylsilyl esters and screened and quantified by high temperature-GC (HTGC). Lipid characterisation will be accomplished by HTGC/MS. This will allow us to determine whether other diagnostic compounds (e.g. plant leaf waxes) are present in addition to animal fats, and to assess the level of preservation (e.g. Fig. 108). Where sufficient concentrations of fatty acids and other acyl lipids are present, these sherds will be base-treated (0.5 NaOH/MeOH) and derivatised to their fatty acid methyl esters (FAMEs) with BF 3 /MeOH (14% v/v) of known isotopic composition. The FAMEs will then further screened by GC, using a polar capillary column and submitted to GC-C-IRMS in order to determine the d 13 C values of the individual fatty acids. The d 13 C values obtained will be corrected for the added derivatising carbon (Jones et al., 1991) and post-industrial revolution effects on atmospheric CO 2 (Friedli et al., 1986) will also be accounted for by the addition of 1.2‰ to the values obtained from the modern reference materials (see below).


Compound-specific stable isotope analyses of reference plant tissues and animal fats

Botanical samples – Analyses of the fatty acid and carbohydrate components of plants likely to have been consumed by ruminants will underpin our interpretations of the animal fat residues from the archaeological pottery (see Fig. 106 above). The d 13 C values will be determined for the major fatty acids (C 16:0 , C 18:0 , C 18:1 , C 18:2 and C 18:3 ) and carbohydrates (glucose, arabinose, xylose, mannose, galactose, rhamnose, inositiol and fucose) from the most significant plants that are known or likely to have been present in antiquity. Expert advice on the key species is being provided by Dr Mark Nesbitt (Royal Botanic Gardens, Kew, London). Suggested species include C 3 grasses ( Aegilpos spp., Avena spp. , Hordeum spp. and Bromus spp. ), C 4 grasses ( Cyperaceae spp. and Chenopodiaceae spp. ) legumes ( Coronilla spp. , Medicago spp. , Trifolium spp. and Trigonella spp. ), leaves ( Quercus spp. ) and crops ( Triticum spp. , and Vicia spp. ). Although C 4 plants do not account for a major part of the botanical record, these plants will be included for completeness. Dr Nesbitt has advised us that collections of modern reference materials are available for sampling both within the UK, and in Ankara and Amman. He has also directed us to experts who can provide advice concerning the archaeobotany of the regions, namely: Dr Mike Charles, Sheffield University and Dr Carol Palmer, Leicester University.

Reference animal fats - We have already built up a very substantial database of d 13 C values for animal fats raised on C 3 diets (see Fig. 106 above). These studies will be extended to include animals raised on more isotopically varied diets. To this end, modern reference fats will be collected from the specific regions to be studied. We also have a collaboration with Dr Matthew Sponhiemer, Departments of Biology, Geology and Geophysics, University of Utah, Salt Lake City, USA, who has agreed to supply various tissues and milk, from representatives of all the major domesticated animals, raised on isotopically controlled diets comprising C 3 and C 4 plants.

Data analysis - The d 13 C 18:0 in milk fat is c. 2.3‰ more depleted than the d 13 C 18:0 in the adipose fats (as discussed above). This phenomenon does not affect the d 13 C 16:0 in either the milk fat or the adipose fat which are in the region of -29.2‰. Therefore the D 13 C values, defined as d 13 C 18:0 - d 13 C 16:0 , can also be used to differentiate between ruminant dairy fats and ruminant adipose fats where the animal has been fed on a pure C 3 diet as would have been the case in Neolithic Britain (Fig. 109). Pilot studies conducted in this laboratory have shown that the D 13 C values of dairy fats obtained from ruminants raised on a mixed C 3 /C 4 diets are not significantly different from the D 13 C values from milk obtained from ruminants raised on pure C 3 diets. Therefore the D 13 C values can be used to distinguish between ruminant dairy fats, ruminant adipose fats and non-ruminant adipose fats even where the diet may comprise mixtures of C 3 and C 4 plants.

The results will enable us to accomplish our objectives: Firstly, to determine the presence or absence of dairy fats which are indicative of the practice of regular milking. This will indicate the date at which such practices began. Secondly, by comparing the proportions of sherds with predominantly milk fats to those containing predominantly adipose fats, it will be possible to determine the relative importance of milk fats and adipose fats at the site. Thirdly, the extent to which C 3 and C 4 cultivars were consumed by animals as forages/fodder in the regions can be elucidated through the d 13 C values of the principal fatty acids. Finally, differences in the dates of first occurrence of dairy fats at these sites should indicate any underlying spatial pattern in the spread of these practices, either from the Near East to Europe, Europe to the Near East, or the roughly contemporary adoption of dairying in both areas.


Figure 109: The d 13 C 18:0 - d 13 C 16:0 (= D 13 C) values for the reference animal fats.   The D 13 C values for the ruminant dairy fats are more depleted than the ruminant adipose fats; the difference in the mean values is c. 2.8‰ which is highly significant (T-test; p<0.0005).

Publication of Results

The results obtained will be of considerable interest to a wide range of international researchers in this field (including archaeologists, palaeobiologists, chemists etc.) and they will be published in high profile international journals aimed at a general readership (e.g. Nature, Science etc.), as well as those with more specialised readership (e.g. Journal of Archaeological Sciences, Archaeometry, Antiquity ). In addition, the results will be presented as oral/poster presentations at conferences such as Archaeological Sciences, Archaeometry and Stable Isotopes in Ecology. We will also produce individual site-specific reports, as we have done with other projects, which are sent to the archaeologists/museum curators for their archives and incorporated as specialist reports for inclusion in excavation publications.



© Çatalhöyük Research Project and individual authors, 2004