Results of the 'round robin' experiment



Results of the 'round robin' experiment

Laboratory a placed the sherd in a plastic dish with 0.5 ml ammonium hydroxide (5%) and sonicated for 5 min. After this, both dish and contents were treated in a rotating mixer for 30 min. About 3 µl of the resulting ammonia solution was transferred into a well in an agarose gel in a Helena's Laboratories ECW electrophoresis chamber and paired with a second well with 3 µl antiserum. A control positive was prepared in a second pair of wells after which an electric current was passed through the gel causing the samples and antiserum to migrate and come into contact. After the run was completed the gel was dried and strained to make positive responses visible. No proteins were detected in the received sherd. They most likely denatured during the boiling of the sample rendering them unidentifiable with the described method.

Laboratory b ground a small piece of the sherd in a mortar and pestle. The resulting powder was transferred into a sterile test tube with 8 ml acetonitrile (ACN). The mixture was sonicated and centrifuged, after which 5 ml was decanted into a second test tube. From this test tube the solvent was evaporated by gentle heating under a stream of nitrogen. The dry residue was taken up in 100 µl ACN and treated with 30 µl MSTFA. About 1 µl of sample was sandwiched in air and injected, proceeded by 1 µl of ACN, into a Varian GC/MS instrument.

Two different pieces of the same sherd were analyzed, using the method of comparing the C18:0/C16:0 and C18:1/C16:0 ratios, yielding noticeable different results. One sample showed a C18:0/C16:0 ratio of 0.72 and a C18:1/C16:0 ratio of 1.15. The second sample showed ratios of 1.03 and 0.32, respectively. This was tentatively interpreted as the residue of some animal product and most likely veal, eggs or goat milk. The observation that the two sherds yielded different results may be due to oxidation of the unsaturated fatty acids.

Laboratory c received the sherd broken into two vertical strips. The upper portion of one vertical strip was selected as the 'time 0' sherd. The remainders of the pieces were stored in an oven at 75°C. After six days the lower portion of the first strip was removed from the oven and stored in a freezer at -20°C. The same was done, after 12 days, with the upper portion of the second strip. Residues were then extracted from the 'time 0' sherd and from those stored in the oven for six and 12 days. Contaminants were removed by grinding off the surfaces after which the sherd was crushed. The resulting powder was mixed with 30 ml of a chloroform and methanol mixture (2:1, v/v) and sonicated (2 x 10 min). Solids were removed by filtering into a separatory funnel. The solvent mixture was washed with 16 ml ultra-pure water and left until it separated into two phases. The lower chloroform-lipid phase was transferred into a flask from which the chloroform was removed by rotary evaporation. Any remaining water was removed by evaporation with 1.5 ml benzene. The dry residue was transferred into a vial with 1.5 ml of the chloroform and methanol mixture and stored. A 200 µl sample of the solution was dried under nitrogen and treated with 6 ml 0.5 N anhydrous hydrochloric acid in methanol. After cooling 4 ml ultra-pure water was added. The fatty acid methyl esters (FAMES) were recovered with 3 ml petroleum ether and transferred into a vial. The solvent was removed by heat under a gentle stream of nitrogen. The dry residue was transferred into a GC vial with 1 ml iso-octane.

Analysis was performed on a Varian 3800 gas chromatograph fitted with a flame ionization detector. Chromatogram peaks were integrated using Varian Star Chromatography Workstation software and identified through comparisons with several external qualitative standards (NuCheck Prep, Elysian, MN). To identify the residue the relative percentage composition was determined first with respect to all fatty acids present in the sample, and second with respect to the ten fatty acids utilized in the identification criteria: C12:0; C14:0; C15:0; C16:0; C16:1; C17:0; C18:0; C18:1ω9; C18:1ω11 and C18:2. Medium chain fatty acids represent the sum of C12:0, C14:0 and C15:0; while C18:1 is the sum of all isomers. It must be understood that the identifications given do not necessarily mean that those foods were actually prepared because different foods of similar fatty acid composition and lipid content would produce similar residues.

Significantly more fatty acids were recovered from the upper portions than of the lower portions of the sherd and, consequently, these provide the best information about the residue. The characterization is based on the relative fatty acid composition of the residue extracted from the upper portion of the second strip. Due to the low level of unsaturated fatty acids in the residues, it is possible to establish decomposition trends after only 12 days of oven storage. The sum of medium chain fatty acids exceeds 30%, while the C18:0 and C18:1 isomer levels are low. By applying published criteria it is possible to say that the partially decomposed residue is typical of decomposed plants. This category includes low fat-content plants such as roots or tubers, greens as well as certain berries and seeds. North American foods known to produce residues with high levels of C14:0 when boiled are biscuit root and Chenopodium seeds.

Laboratory g ground off about 0.1 g of the internal surface of the sherd was ground off, mixed the resulting powder with 1 ml of a chloroform and methanol mixture (2:1, v/v), sonicated and centrifuged. The supernatant was decanted and dried under a stream of nitrogen. The dry residue was treated with 30 µl BSTFA and about 1 µl of the derivatized sample was injected into a Hewlett Packard 5972 CG/MS.

Next to a trace of C12:0, the sample appeared to contain a number of free fatty acids: C14:0, C16:0, C18:0 and C18:1 as well as the methyl-esters of C16:0 and C18:0. Neither short chain fatty acids nor polyunsaturated fatty acids were identified. No attempt was made to determine the ratios of the free fatty acids but C16:0 appeared predominant. There were traces of odd carbon number fatty acids, but these were in low abundance. Cholesterol was present in abundance along with cholesterol oxidation products and a trace of squalene. No phytosterols or alkanes were found. No terpenoids and no wax esters were identified, indicating that the source most likely did not contain a resin or a wax. On the basis of these results it was suggested that the residue is of animal origin, the absence of short chain fatty acids pointing to a non-dairy source like veal or eggs.

Laboratory h ground the sherd in a mortar and pestle. Lipids were extracted in a chloroform and methanol mixture by sonification. This was centrifuged and decanted after which the solvents were evaporated under nitrogen. The dry residue was treated with BSFTA, saponified with sodium methoxide in methanol and methylated with a BF₃-methanol complex. The sample was then analyzed in a GC-17A Shimadzu Chrom-Perfect gas chromatograph, with a 30 m Agilent DB-5 column raised 5°C/min from 40°C to 280°C, after a split injection. No activity was found in the received sherd.

Laboratory j removed the outside surfaces of the sherd, to avoid possible contaminants, after which a total of 12 samples were obtained at evenly spaced intervals from the rim to the base of the vessel segment, six from the interior and six from the exterior. Aliquots of the ceramic powder, weighing an average of 4.8 mg, were manually compressed in tin foil capsules and combusted in a Carlo Erba elemental analyzer interfaced with a MAT 252 isotope ratio mass spectrometer. δ¹³C and δ¹⁵N values and weight% C and N were measured for each sample.

Carbon concentrations averaged 2.69% on the vessel interior and 1.65% on the exterior. Vessel interior δ¹³C values were systematically more negative than exterior values. The average δ¹³C for the interior was -22.3‰, that for the exterior was -19.2‰. Interior and exterior δ¹³C values varied greatly with distance from the rim, first increasing by almost 5‰, then decreasing by 4‰ and then systematically increasing toward the base. Interior and exterior values co-varied systematically with distance below the rim.

Nitrogen concentrations in the original samples appeared too low for detection in all vessel exterior samples and in the three interior samples closest to the rim. Amounts of nitrogen in the three lowest interior samples (>22 mm below the rim) increased systematically from 0.16 to 0.21% with distance below the rim. Their average δ¹⁵N value was 5.08‰, but these nitrogen samples were too small for accurate isotopic analysis. The sample weights were doubled (averaging 10.4 mg) to obtain more reliable estimates of nitrogen concentrations and for reliable δ¹⁵N values. Nitrogen concentrations were 0.05 to 0.06% in the three samples closest to the rim, and increased systematically from 0.05% to 0.20% between 16 and 41 mm below the rim. Atomic C/N ratios increased slightly and then decreased away from the rim. The three lowest samples contained enough nitrogen for reliable isotopic analysis. Their δ¹⁵N values averaged 6.00‰.

Co-variance of interior and exterior δ¹³C values suggests that the vessel contents were absorbed throughout the thickness of the vessel. The difference in δ¹³C between the interior and exterior suggests that the vessel's cooked contents had a lower δ¹³C value than that of the fuel used to fire the vessel. Alternatively, if the vessel had been 'seasoned' by the potter, such as by boiling with milk, then this pattern could be due to application of a substance with a δ¹³C value that differed from that of the fuel used in firing, to seal either the interior or the exterior of the vessel. The distinct deflection to the most negative δ¹³C values at 21-22 mm below the rim may indicate the 'boil line'. This hypothesis is based on the fact that fats have a δ¹³C value approximately 5‰ more negative than those of other biochemical fractions of foods (proteins and carbohydrates). If the food was cooked in liquid, then isotopically lighter fats would float to the surface, and would be absorbed at higher levels than heavier foodstuffs. High C/N ratios near the rim are also consistent with this hypothesis as fats have extremely high carbon contents and lack nitrogen. The increase in nitrogen content and decrease in atomic C/N ratios beginning 22 mm below the rim indicate protein-rich food components were absorbed in lower parts of the vessel.

The increase in δ¹³C values between the rim and 10 mm below the rim is unlikely to be due to the isotopic composition of the food cooked in the vessel. Perhaps the vessel was fired upside down, with the rim in contact with fuel of C3 origin. Alternatively, if the pot was seasoned with a 13C-enriched substance, the rim was not treated. Increasing δ¹³C values beneath the boil line suggest that a mixture of C3 and C4 foodstuffs or proteins of a mixed-feeding animal were cooked in this pot. The δ¹⁵N values are high enough to suggest the presence of meat, milk or blood of an herbivore, perhaps combined with plant foods. The systematic patterns of carbon and nitrogen concentrations and isotopic composition of the interior and exterior of this vessel were entirely unanticipated. The explanations for these patterns suggest that a simple sampling strategy and method of analysis provides a considerable amount of useful evidence about foods cooked in porous ceramic vessels. Variation in isotopic composition within a vessel used to cook a single meal can be substantial. Therefore control over sample position within the vessel is necessary for adequate interpretation of carbon and nitrogen elemental and isotopic variation of absorbed organic matter in potsherds.

Laboratory k cut the sherd into four pieces, two of which were crushed into a fine powder using a mortar and pestle. Lipids were extracted from 500 mg of the powder in 2 ml chloroform/methanol (2:1, v/v) by sonification after which the solvents were evaporated under a stream of nitrogen. The dry residue was then treated with BSTFA with 1% TMCS. Lipids in another 500 mg were saponified with 0.5 ml 2N KOH, acidified with HCl and extracted with diethyl ether. The solvent was then evaporated under a stream of nitrogen and the dry residue treated with BSTFA with 1% TMCS. The derivitized residues were analyzed in a Hewlett Packard 5890 series II gas chromatograph (HTGC) and a Hewlett Packard 5890 gas chromatograph with a 5972 mass selective detector (GC/MS).

The analysis of the total lipid extract, by HTGC, showed the presence of triacylglycerols with a total carbon number between 44 and 56, the main constituent having 52. After saponification and GC/MS, the main fatty acids appeared to be C16:0, C14:0, C18:0, C12:0, C18:1 and C16:1. The C16:0/C18:0 ratio for one part of the sherd was 3.175 by HTGC and 4.847 by GC/MS. For the other part of the sherd the C16:0/C18:0 ratio was 3.09 by HTGC. The distribution of triacylglycerols indicates a residue of animal origin, a C16:0/C18:0 ratio of around 3 may be explained by the presence of egg. It was therefore concluded that an animal product was the source of the residue.

The source of the residue in X-11 was fresh camel milk, obtained in January 2004 at the camel market at Daraw, between Aswan and Luxor in Upper Egypt. About 200 ml. of this was mixed with an equal amount of mineral water and stored in a new, ungazed ceramic vessel (a 'tagen') purchased in Luxor. The whole was wrapped in aluminum foil and allowed to sit for 24 hours.

The next day, the vessel and its content were put in an oven and boiled for an hour. After cooling down this was repeated once, after which the vessel was left untouched for 24 hours. It was then unwrapped, rinsed with cold water and stored. In May 2004 the vessel was machine-cut into twelve pieces after which the eleven rim sherds were send to laboratories interested in participating in the experiment.

Seven laboratories forwarded their results in time to be included in this report, before the source was revealed during the 70th Annual Meeting of the SAA (Salt Lake City, 31 March 2005). Although the residue in the ceramic matrix will be different from the composition of fresh camel milk, this is given here (courtesy of the FAO) as an indication of what could have been expected.

Hans Barnard (UCLA) and Jelmer Eerkens (UC Davis)
c/o PO-box 951511; Los Angeles, CA 90095

Your suggestions, additions, comments and corrections are welcome and your participation even more so.