Journal of Archaeological Science 34 (2007) 804e813 http://www.elsevier.com/locate/jas Paleodietary implications from stable carbon isotope analysis of experimental cooking residues John P. Hart a,*, William A. Lovis b, Janet K. Schulenberg c, Gerald R. Urquhart d a Research and Collections Division, New York State Museum, 3140 Cultural Education Center, Albany, NY 12230, USA Department of Anthropology and MSU Museum, 354 Baker Hall, Michigan State University, East Lansing, MI 48824, USA c Division of Undergraduate Studies, 129 Grange Building, The Pennsylvania State University, University Park, PA 16801, USA d Lyman Briggs School of Science, E-194 Holmes Hall, Michigan State University, East Lansing, MI 48825, USA b Received 18 May 2006; received in revised form 24 August 2006; accepted 25 August 2006 Abstract The regional timing of maize introduction in eastern North America is a long-standing topic of archaeological interest. Most recently, Morton and Schwarcz [2004. Paleodietary implications from stable isotopic analysis of residues on prehistoric Ontario ceramics. Journal of Archaeological Science 31, 503e517] investigated the timing of maize introduction in Ontario through isotope analysis of charred cooking residues adhering to the interior of prehistoric ceramic containers. They interpret their results to suggest maize was incorporated into diets after A.D. 600. We assess their approach and conclusions with stable carbon isotope assays on three sets of experimental cooking residues, evaluating the variable combustion of carbon fractions, contributions of fats and carbohydrates, and the contribution of total carbon. We also undertake multiple resource modeling of two part food mixes with green maize and maize flour. Our results suggest that systematic under representation of maize can result depending on residue composition and that some prior knowledge of C3 plant and animal contents is necessary to interpret stable carbon isotope values on cooking residues. We question the independent use of stable carbon isotope analysis of charred cooking residues as a viable technique for extracting paleodietary information. Ó 2006 Elsevier Ltd. All rights reserved. Keywords: Prehistoric cooking residues; Carbon isotopes; Prehistoric cooking techniques 1. Introduction The timing of maize’s (Zea mays ssp. mays) introduction into various regions of eastern North America (ENA) and its adoption as a dietary staple continue to be long-standing topics of interest among archaeologists and paleoethnobotanists [17]. Various sources of information have been used to reconstruct the histories of maize in ENA including direct AMS dates on macrobotanical remains [4,6]; stable carbon isotope analysis of human bone [3,15], charred cooking residues [27,31,44], and absorbed organic pottery residues [39e41]; pollen * Corresponding author. Tel.: þ1 518 474 3895; fax: þ1 518 486 2034. E-mail addresses: [email protected] (J.P. Hart), [email protected] (W.A. Lovis), [email protected] (J.K. Schulenberg), [email protected] (G.R. Urquhart). 0305-4403/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.jas.2006.08.006 recovered from sediments [10,55]; and phytoliths recovered from cooking residues [19,47,48]. Recently, several studies have used stable carbon isotope analysis of food residues from ceramic vessels to investigate the adoption of maize in specific regions, and to investigate the ways maize was incorporated into paleocuisine [31,39,44]. In the largest of these studies, Morton and Schwarcz [31] investigated the timing of maize’s introduction in Ontario, Canada using stable carbon isotope analysis of charred cooking residues adhering to the interiors of prehistoric pottery sherds. Based on analysis of 137 residue samples from 50 components of 45 sites dating between approximately 680 B.C. and A.D. 1725, Morton and Schwarcz suggest that maize was incorporated into diets after A.D. 600, consistent with isotopic evidence from human bone [23,45] and macrobotanical evidence [6]. However, their results suggest only a small J.P. Hart et al. / Journal of Archaeological Science 34 (2007) 804e813 .maize was apparently a minor ingredient in the mixtures actually cooked in the pots. The remainder of dietary maize must then have been consumed in different forms, either as whole grains or cobs, or in some prepared form. Even if the residues are an accumulation from many meals and therefore an average value of the food eaten, we cannot account for their low mean d13C value given the clear evidence that maize was a major food for the native peoples. [31, p. 515] Morton and Schwarcz’s interpretations are based on a formula used to estimate the percentage of C4 plants (PC4) that contributed to the creation of a residue. This formula assumes a linear relationship between d13C and the proportion of maize, the most likely C4 plant used in ENA, relative to C3 resource contribution to the residue. In this article, we assess Morton and Schwarcz’s approach by employing the results of stable carbon isotope assays on experimental cooking residues. We use the results from three independent experiments using 10% increments of maize and wild rice (Zizania sp.) [19], maize and chenopodium (Chenopodium album) [44], and maize and white-tailed deer (Odocoileus virginianus) meat (reported here for the first time), to assess the assumption of linearity between d13C and maize proportions in charred, encrusted cooking residues. Our results indicate that the relationship is not always linear, and as a result, the PC4 formula can substantially over- and underestimate maize proportions (see [25,35] for similar results in animal tissues). We test three hypotheses that may account for the patterning evident in the experimental residue d13C values relative to maize proportions. Total carbon content of the resources used in creating the experimental residues provides the best-fit for those patterns. We then assess the effectiveness of Morton and Schwarcz’s threshold for identifying maize’s presence in a residue. Our results suggest that prior knowledge of the C3 plants and animals consumed on and cooked together at a site is necessary in order to interpret stable carbon isotope values on prehistoric charred cooking residues. The results also suggest the importance of prior knowledge of food preparation and preservation techniques, especially whether green (wet) or mature (dry) maize was cooked in the pot. In sum, we conclude that d13C values on charred cooking residues cannot be used to infer either presence of maize or its percent contribution to a specific residue. advocated caution in using bulk stable carbon isotope analysis as the basis for interpreting the presence or absence of maize in a residue. Similar to Morton and Schwarcz [31], Schulenberg [44] attempted to delineate a minimal isotopic threshold for the presence of maize in archaeological residues. However, her experimental results suggested that direct interpretation of d13C is problematic. Hart et al. [19] combined d13C analysis with phytolith analysis of charred cooking residues from an assemblage of central New York sherds. The results of this study suggest that depleted d13C cannot be used to indicate the absence of maize in a residue; residues with depleted d13C contained phytolith assemblages identified as maize [19]. Morton and Schwarcz’s [31] recent application of this technique to a large sample of vessels from Ontario represents an important effort to standardize the interpretation of d13C through a mathematical model. Morton and Schwarcz [31] use the following formula to estimate the amount of maize present in a residue: PC4 ¼ ½ðds d3Þ=ðd4 d3Þ 100 ð1Þ where d3 is the mean d13C for C3 plants (28&), d4 the mean d13C for C4 plants (9&), and ds the d13C value for the residue. The formula assumes a linear relationship between the proportion of maize contributing to a residue and the residue’s d13C value. We obtained d13C on three series of experimental residues with known proportions of maize and a second, C3 resource. The combined results allow us to test the underlying assumption of linearity (Fig. 1). In our assessment of the Morton and Schwarcz formula, it was necessary to recognize several important aspects of their work. They note, ‘‘the equation assumes that each sample is a mixture of only C3 and C4 plants, whereas herbivore and fish flesh were also present in the [prehistoric] diet’’ [31, p. 508]. We attempt to directly address the impact of C3 herbivores in one of our experimental trials, although direct comparison with Morton and Schwarcz’s experimental data was -5 0 10 20 30 40 50 60 70 80 90 100 -10 -15 13C amount of maize was cooked in the vessels, even during times when it is known that maize was a substantial dietary component. Morton and Schwarcz [31] interpret the apparent under-representation of maize in the cooking residues as an indication that maize was processed in other ways: 805 -20 -25 -30 2. The PC4 formula Two pilot studies in ENA investigated the applicability of stable carbon isotope analysis to food residues [39,44]. While these experimental applications varied in focus and technique (Reber’s [39] focused on absorbed residues while Schulenberg’s [44] focused on encrusted residues), both researchers -35 Percent Maize Fig. 1. Comparison of experimental results with PC4 model. Plots of the experimental residues are based on the weight values in Table 1. (X ¼ deer meat and maize flour, - and : ¼ wild rice and maize flour, B ¼ chenopodium and frozen maize kernels.) 806 J.P. Hart et al. / Journal of Archaeological Science 34 (2007) 804e813 not possible because they do not specify proportions for the residues made with more than one resource. They discuss the potential impact of ‘‘variations within plant species and individual plants’’ [31, p. 508], although unlike our replications, their experimental work only included the common bean (Phaseolus vulgaris) as a C3 plant [31, p. 513]. For our purposes, Morton and Schwarcz [31, p. 515] pose a significant question about their results: ‘‘Why is there so little evidence of the presence of C4 plants. in these residues?’’. They argue that maize must have been a small contributor to the meals cooked in pots. To explain the discrepancy between the isotope evidence from encrusted food residues and those of measurement on human bones of the same time periods, Morton and Schwarcz [31, p. 515] suggest that maize was consumed primarily on the cob or as whole kernels. The interpretation of the stable carbon isotope values highlighted by this discrepancy is the central question to which we directed our experimental data and analysis. We believe our results provide a compelling alternative explanation to the ones proposed by Morton and Schwarcz. 3. Materials and methods Three series of experimental residues were created using mixtures of maize with one other food source. In the first experiment, Schulenberg [44] used mixtures of supermarketpurchased frozen sweet corn and chenopodium (C. album) harvested in southeastern New York to create a series of 11 residues. The experimental mixtures were done in 10% increments by weight. Whole chenopodium seeds and maize kernels were cooked in water in low-fired clay pots over coals. The water was allowed to evaporate and the contents to burn. The resulting charred material was then ground in a mortar with a pestle. Three subsamples of each residue were submitted to Geochron Laboratories for carbon isotope assay. The second experiment, conducted by Hart et al. [19], combined Iroquois White Flour maize, grown in New York by Jane Mt Pleasant of Cornell University, and wild rice obtained from a commercial vendor in Minnesota. Dried wild rice seeds and maize kernels were ground separately into flour and sieved through 1-mm mesh to remove any large particles. Eleven, dry 1.25-ml mixtures were created in 10% increments by volume. These mixtures were simmered in 100 ml of water for 20 min with occasional stirring. The mixtures were then decanted into cans and placed on a wood fire. The water was allowed to evaporate and the mixture to burn. Two subsamples of each residue were submitted to the Illinois State Geological Survey (ISGS) Isotope Geochemistry Lab for carbon isotope assay. The third experiment, reported here for the first time, was conducted by Lovis using lean white-tailed deer meat and Iroquois White Flour maize from the same batch used in experiment 2, albeit from a different ear. The deer meat came from the upper peninsula of Michigan in a non-agricultural region and where maize is not used as bait, thereby minimizing the potential for C4 plant consumption. Isotopic analysis of prehistoric deer bone suggests maize was not a dietary component of at least some deer populations [21,54] unlike modern deer populations [5]. The same procedures were followed in this experiment as related for experiment 2. The deer meat was frozen before grinding and then weighed prior to mixing. In order to thoroughly mix the deer meat with the maize flour it was necessary to place the mixtures in a blender. One subsample of each mixture was submitted to ISGS for stable carbon and nitrogen isotope assay. During boiling and carbonization, the maize/meat mixture separated into both ‘‘light’’ and ‘‘heavy’’ fractions. The former was composed largely of separated plant material suspended in a froth (probably high in starches/carbohydrates per Pierce [36, p. 138]) and some fats, and the latter primarily of denser meat particles. This resulted in carbonized residues in two locations: at the upper part of the fluid line, and at the base of the container (see Hastorf and DeNiro [20] for sampling of basal sherds). In this experiment, the two residues were combined. In future research, it might be productive to sample each residue location separately. It is possible that the locations from which residues are drawn may affect isotope composition and therefore outcomes, i.e. it is expectable that basal residues may register higher C3 content, and jar wall residues higher C4 content. 4. Results Results of the d13C assays on the experimental residues are presented in Table 1 and Fig. 1. Clearly, the underlying assumption of linearity in the PC4 formula is not realized by these results; the relationship between d13C and proportions of C3 and C4 plants/consumers contribution to a residue in our experiments is distinctly nonlinear. The non-linearity varies depending on the products used to create the samples. The maizeedeer residues produce a convex curve, while the maizeewild rice and maizeechenopodium residues produce concave curves. The maizeechenopodium curve is especially pronounced. 5. Modeling d13C We tested three hypotheses to account for the nonlinear relationship between maize proportions and d13C values: 1. Fractionation, or the uneven combustion of carbon isotopes where the reaction rates of the heavier 13C and lighter 12C differ, produces a nonlinear relationship between the proportion of maize and the d13C values. 2. Different contributions of fats and carbohydrates (which burn at different temperatures) to the residues create the nonlinear relationship. 3. Uneven contribution of total carbon from each carbon source creates the nonlinear relationship. Fractionation of stable carbon isotopes is mass dependent [14]. The square root of masses gives the fractionation ratio [14]: rffiffiffiffiffiffiffi rffiffiffiffiffi m13 13 Ratio ¼ ¼ ¼ 1:04 12 m12 ð2Þ J.P. Hart et al. / Journal of Archaeological Science 34 (2007) 804e813 807 Table 1 Isotope results for experimental residues Maize-wild rice Maize-deer 13 13 13 Maize-chenopodium % Maize (vol.) % Maize (wt.) d C-1 d C-2 d C-avg % Maize (vol.) % Maize (wt.) d C % Maize (wt.) d13C 0 10 20 30 40 50 60 70 80 90 100 0 7.8 15.9 24.5 33.6 43.1 53.2 63.4 75.2 87.2 100 26.3 24.0 24.2 23.3 22.8 22.3 22.5 18.7 16.3 14.7 11.8 25.4 25.6 24.5 24.1 22.1 21.5 20.0 20.4 18.3 14.3 11.9 25.9 24.8 24.4 23.7 22.5 21.7 21.3 19.6 17.3 14.5 11.9 0 10 20 30 40 50 60 70 80 90 100 0 5.3 12.4 19.8 26.3 38.2 44.3 57.4 71.6 84.6 100 24.5 21.8 19.7 17.2 16.4 16.4 15.6 13.7 13.0 12.6 12.7 0 10 20 30 40 50 60 70 80 90 100 32.4 32.7 32.6 31.6 31.8 30.5 28.2 28.8 25.6 22.9 14.3 This means that 1.04 times as much of the original 13C compared to the original 12C remains in residues (relative to the original amounts): Ratio : 13Coriginal 13Cfinal 12Coriginal 12Cfinal ¼ 1:04 13Coriginal 12Coriginal = ð3Þ As a result, fractionation results in a linear relationship, inconsistent with our experimental data. The hypothesis that fractionation affects the d13C ratios of the cooking residues can, therefore, be rejected. Fats and carbohydrates burn at different temperatures [9], and as a result, higher proportions of the carbohydrates in a food will char relative to the fats (lower burning temperature). However, to produce the curves resulting from our experimental residues, there would need to be a large difference in the amount of fats vs. carbohydrates in the resources used to create them. Percent fat and carbohydrate data for the resources used in the experimental residues are presented in Table 2. Deer has a much higher fat/carbohydrate ratio than maize. This could have resulted in the heavier d13C values even when maize is a low fraction (5% or 10% of diet), producing the convex curve in Fig. 1. However, the chenopodium and wild rice fat/carbohydrate ratios are similar to that of maize. The differences in the ratios between maize and the other resources are not sufficient to produce the Table 2 Overall carbon content of maize flour, green maize (kernels), chenopodium, and deer meat Maize flour Green maize (kernels) Chenopodium Wild rice Lean deer meatdfrozen Fat (%) Carbohydrate (%) Protein (%) Percent of mass from carbon 4.9 1.2 6.3 1.2 7.1 79.1 19.0 45.0 81.2 0.0 13.4 3.2 28.8 16.0 21.8 44.2 10.6 39.2 43.5 17.2 Fat, carbohydrate and protein values obtained from USDA Nutrient Database figures, http://www.nal.usda.gov/fnic/foodcomp/search/. Percent of mass from carbon calculated from formula carbon ¼ 0.80 (fat) þ 0.42 carbohydrate þ 0.53 protein. 13 strongly curved relationship shown in Fig. 1. As a result, this hypothesis is also rejected. To address the possibility that the total carbon in the various resources affected the d13C of the experimental residues, we used percent of mass from carbon (from fats, carbohydrates, and protein) to calculate how much carbon each would contribute if equal original weights were used to create the mixtures. Fatty acid mass is w80% carbon [32], carbohydrate mass w42% carbon, and protein mass w53% carbon [33] (Table 3). Percent of mass from carbon (%C) was calculated using Eqs. (4) and (5); carbon contents for mixtures were calculated using Eq. (6) (Table 3). %CMaize ¼0:80FatMaize þ0:42CarbsMaize þ0:53ProteinMaize ð4Þ %COther ¼0:80FatOther þ0:42CarbsOther þ0:53ProteinOther ð5Þ where FatMaize, CarbsMaize and ProteinMaize are, respectively, the grams of fat, carbohydrates and protein amounts per 100 g of maize. Frozen maize and maize flour had different values. Similarly FatOther, CarbsOther and ProteinOther are the fat, carbohydrates, and protein amounts per 100 g of the other substance. The combined mixtures’ d13CM values were calculated from the following mixing equation [9]: d13 CM 13 d CMaize ð%Maize %CMaize Þ þ d13 COther ð%Other %COther Þ ¼ ð6Þ ð%Maize %CMaize Þ þ ð%Other %COther Þ where %Maize is the percentage of maize in the mixture (0, 10, 20, . 100), %Other is the percentage of the other material in the mixture (1 %Maize), %CMaize is the percentage of maize Table 3 Biomolecules and their carbon content Source Biomolecule % Carbon Fatty acids Carbohydrates Proteins CnHnO C6H12O6 (ReCH(NH2)eCOOH)n w80.0 42.0 53.0 J.P. Hart et al. / Journal of Archaeological Science 34 (2007) 804e813 by weight that is carbon (Eq. (4)), %COther is the percentage by weight of the other that is carbon (Eq. (5)), d13CMaize is the initial d13C value for maize, and d13COther is the initial d13C value for the other substance in the mixture. For each model, we ran the equation using the observed d13C values from the particular experiment to generate a best-fit approximation (Fig. 2). The Green MaizeeChenopod model used 13.7 for maize and 32.4 for chenopod; the Maize FloureWild Rice model used 11.87 for maize and 26 for wild rice, and the Maize FloureDeer Meat model used 12.6 for the maize and 24.49 for the deer meat. The models also incorporate maximum and minimum d13C values for the substances combined. Reported d13C values for maize range from 8.7 to 15.7 [2,3,50], for chenopod from 26 to 33.8 [43,44], for wild rice from 26 to 30.9, and for deer meat from 23 to 29.5 [49]. These values created the outer bounds for our model. Modeling the experiment using frozen maize and chenopodium, we find that maize kernels have high water content, resulting in a reduced amount of carbon (10.6%) relative to the chenopodium (39.2%). Following Eq. (6), this difference in contribution of carbon to the residues produces a concave curvilinear relationship (Fig. 3). The model for the experiment that used dry maize flour and dry wild rice flour had nearly equal amounts of carbon in both substances. The carbon content of dry maize flour is 44.2%, while dry wild rice flour contributes 43.5% carbon by weight (Table 3). This contribution produces a nearly linear relationship (Fig. 4). Finally, the model based on the dry maize flour and frozen deer meat experiment produced a convex curvilinear relationship because the carbon content of dry maize flour is 44.2%, while frozen deer meat would contribute 17.2% carbon by weight, (Table 3, Fig. 5). In sum, the total carbon in resources used to create the experimental residues produces curves (Fig. 4) notably similar to the curves derived for the experimental residues (Fig. 1). Departures of the experimental residue d13C values from the predicted values may reflect changes that occur during the -5 0 10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90 100 Model Minimum Maximum Observed -15 -20 -25 -30 -35 -40 Percent Maize Fig. 3. Maximum, minimum, modeled and observed for wet/green maize and chenopodium. charring of vegetal matter as has been noted in laboratory experiments [37,51], or the loss of starches (carbohydrates) as a consequence of poor heat control and boil over [36, p. 138] among other uncontrolled variables. As a result, this hypothesis cannot be rejected. To approximate more realistic situations where multiple products would have been cooked in a single pot, we modeled a slurry composed of one third each of deer, chenopodium and wild rice. This slurry had a calculated d13CSlurry of 29.4, with a minimum of 32.91 and maximum of 27.2 (Eq. (7)) and carbon content of 33.3 g/100 g (Eq. (8)). Using Eq. (6) we calculated d13CM values for a mixture of the slurry with wet 0 -5 0 -10 13C 808 100 0 10 20 30 40 50 60 70 80 90 100 Model -5 Minimum Maximum -10 -10 Observed -15 13C 13C -15 -20 -20 -25 -25 -30 -30 -35 Percent Maize 13 -35 13 Fig. 2. Model d C values based on Eq. (6) and average C3 and C4 d C values. (- ¼ wild rice and maize flour, : ¼ wild rice and wet maize, B ¼ chenopodium and wet maize, X ¼ deer meat and maize flour.) Percent Maize Fig. 4. Maximum, minimum, modeled and observed for dry maize flour and wild rice flour. J.P. Hart et al. / Journal of Archaeological Science 34 (2007) 804e813 0 0 10 20 30 40 50 60 70 80 90 100 Model -5 Minimum Maximum Observed 13C -10 -15 -20 -25 -30 -35 Percent Maize Fig. 5. Maximum, minimum, modeled and observed for dry maize flour and deer meat. (green) maize and maize flour, with maize ranging from 0% to 100% in 10% increments and slurry comprising the remainder. for all but a few of the residues. Use of the PC4 formula to identify the proportion of maize contributing to a prehistoric cooking residue, then, is questionable without prior knowledge of the C3 plants/consumers cooked in the pot, if maize was also cooked, and if so whether the maize was cooked green (wet) or mature (dry). Morton and Schwarcz [31] use d13C values of 24& to represent the presence of maize, and values <28&, to indicate no C4 plant contribution to the residue. Using an average d13C value of 26& for industrial era C3 plants, results in a threshold for determining a contribution of maize of 22& using the 4& offset suggested by Morton and Schwarcz. The threshold, requires an approximate contribution of 35e55% maize in the maizeewild rice residues, approximately 5% for the deeremaize residues, and 90% for the maizeechenopodium residues (Fig. 7). Given the varied results of our experiments and models, which are dependent on the mix of resources, any threshold chosen to determine the presence of maize in charred residues is likely to result in errors. These results further demonstrate the problematic nature of the PC4 formula and its application and emphasize the variation in d13C values depending on resources contributing to a residue. 13 d CDeer %CDeer þ d13 CChenopod %CChenopod þ d13 CWildRice %CWildRice d CSlurry ¼ ¼ 29:4 %CDeer þ %CChenopod þ %CWildRice 13 %CDeer þ %CChenopod þ %CWildRice %CSlurry ¼ ¼ 33:3 3 809 ð7Þ 7. Discussion ð8Þ The results of this modeling, which are also curvilinear, suggest that the d13C value of the residue depends heavily on whether or not the maize used was green (wet) or mature (dry) (Fig. 6). 6. Identifying maize in prehistoric cooking residues The total amount of carbon contributed to a residue by a resource, not strictly the proportion of the resource in the food being cooked, determines the d13C value of the residue. This means that the PC4 formula has the potential to substantially under- or over-estimate the percentage of maize in the food cooked in a pot. Because the direction of the error depends on the nature of the food products contributing to the residue, there is no simple correction factor for this problem. The percent of maize contributing to the experimental residues is plotted against PC4 values in Fig. 1. In this figure, we use industrial era average C3 and C4 d13C values of 26& and 12&, respectively [26, p. 462]. As is evident, the PC4 formula significantly underestimates the percentage of maize We recognize the complexity of the problem that we are collectively addressing. Morton and Schwarcz’s attempt to systematize the way in which we can interpret isotope values from ceramic cooking residues particularly with respect to maize content is a positive step in the process of addressing the problem. However, the PC4 formula is based on the assumption that each resource present in a mixture contributes carbon to the residues identical to that resource’s proportion in the food being cooked. To the contrary, we expect different resources to contribute different amounts of carbon because they differ in carbon content and/or because they release different amounts of carbon during cooking that results in the contribution to the residue (e.g., one resource breaks down more quickly than another), or because they have been cooked dried (e.g., matured and stored) or green. This means that at least three sources of variation can potentially result in a broad range of isotope combinations not accommodated by the PC4 formula. Our experimental data clearly demonstrate this, and the results suggest that major interpretive errors related to maize content may occur and are resource mix dependent. As a result, Morton and Schwarcz’s threshold for maize presence can potentially result in erroneous determinations of maize’s contribution to cooking residues. J.P. Hart et al. / Journal of Archaeological Science 34 (2007) 804e813 810 a) Dry Maize 0 0 10 0 20 30 40 50 60 70 80 90 30 40 50 60 70 80 90 100 -10 Minimum -15 Maximum 13C -10 13C 20 100 Model -5 10 -5 -20 -15 -25 -20 -30 -25 -35 -30 -35 Percent Maize Fig. 7. Comparison of experimental values to a threshold for maize presence of 22&. Plots of the experimental residues are based on the weight values in Table 1. (X ¼ deer meat and maize flour, - and : ¼ wild rice and maize flours, B ¼ chenopodium and frozen maize kernels.) Percent Maize b) Green Maize 0 0 10 20 30 40 50 60 70 80 90 100 Model -5 Minimum Maximum 13C -10 -15 -20 -25 -30 -35 Percent Maize Fig. 6. Model d13C values of mixtures of hypothetical slurries of equal parts deer, chenopod, and wild rice mixed with increments of (a) dry maize and (b) green (wet) maize. The most critical problem faced in the interpretation of d13C values is the requirement for knowledge of what resources contributed to each residue formation (see also [25,35]). This requires the generation of independent lines of evidence beyond that of macrobotanical remains. Certainly, macrobotanical remains can provide information about the suite of resources potentially contributing to residues, but unless embedded in residues, any given resource cannot be assumed to have contributed to that residue. Phytolith assemblages may provide more direct evidence of the botanical content of residues, and have been found to be at odds with data generated by analysis of macrobotanical remains [19,48]. This is clearly one avenue by which to clarify residue composition, although not all plants known to have been consumed prehistorically in eastern North America produce phytoliths in their consumable parts. Gas chromatographic methods to identify lipids can provide another source of information on both plant and meat food sources at a site [8,28,38,40,41], although resource specific identifications are not always possible. Additional factors that need to be taken into account include the effects of charring or carbonization on isotopic fractionation. Heating and charring experiments of vegetal and animal resources have shown that changes occur in bulk d13C values [22,37,51]. In one study these changes ranged from 0.5& to þ2.0& at varying temperatures for a single resource [37] and in another from þ1.0& to 1.2& for varied resources cooked together in a stew [22]. A third study showed changes of up to 1& in plant cellulose following various cooking and charring procedures [29]. Although such changes may not have significant affects on paleodietary isotopic analysis of human bone [22,29], they probably account in part for the departures from expected of our experimental residues from the models. Other experiments have shown a potential loss of starches (carbohydrates) as a result of frothing and boil over [36, p. 138]. The separation of fats from other carbon sources during cooking may also result in depleted d13C values in residues formed on vessel necks and rims [27]. The various biochemical components of food resources may vary in their carbon isotope values [50]. Differential preservation of these components during charring may affect the isotope values of residues, although one set of experiments suggested otherwise [50]. This study [50] also found no statistically significant difference between bulk and mass balance calculated carbon isotope values using the isotope values for each biochemical component. We have not systematically addressed these issues. There are also variations in d13C values in individual plant species (see Table 4). The highly depleted d13C values of the chenopodium used in Schulenberg’s [44] experiments, J.P. Hart et al. / Journal of Archaeological Science 34 (2007) 804e813 811 Table 4 Variation in d13C values for selected prehistoric crop remains in eastern North America Plant d13C Range (&) n Source(s) Sumpweed (Iva annua) seed 23.4 to 28.5 11 Squash (Cucurbita pepo) seed Uncharred Squash (Cucurbita pepo) seed 24.9 to 26.4 23.6 to 26.6 3 4 Squash (Cucurbita pepo) rind 23.7 to 28.3 6 Bean (Phaseolus vulgaris) seed Maize (Zea mays ssp. mays) kernel Maize (Zea mays ssp. mays) cob 23.3 to 28.6 8.4 to 15.1 7.4 to 12.6 34 35 5 Gayle Fritz, personal communication, 2004a; Mary J. Adair, personal communication, 2004b Mary J. Adair, personal communication, 2004b [34]; Monaghan, Lovis and Egan-Bruhy, unpublished datac Mary J. Adair, personal communication, 2004b; Hart unpublished datad; Monaghan, Lovis and Egan-Bruhy unpublished datac [18] [18,19,24,53]; Gayle Fritz, personal communication, 2004a [18] All samples are charred unless otherwise noted. a Values from Ozark rockshelters in Arkansas (see corresponding dates in [11]). b Values from various Great Plains sites (see corresponding dates in [1]). c Values from Michigan sites (see corresponding dates in [30]). d Value from Pennsylvania site (see corresponding date in [16]). as well as those from the charred prehistoric residues of unknown composition reported by Lovis [27], suggest that the range of d13C values must be broadened to incorporate such variation. We have incorporated this information into our models. Such highly depleted values are normally associated with either forest floor materials or wetlands [52]. The physical locations of archaeological sites as well as the habitats within which plants such as chenopodium are found make such depleted d13C values more likely than not. Precontact archaeological sites are often situated adjacent to riverine or lacustrine wetlands [7,42]. Furthermore, the disturbed soil situations in which plants such as chenopodium thrive could well be found on the margins of habitation site clearings rather than in the habitation area itself, situations where forest or woodland soils would be present, as well as on disturbed floodplain soils [12,13,30]. We therefore suggest that modeling exercises must reflect this range of variation. If it were possible to identify all subsistence resources used on a site as well as the habitats in which plant resources were grown, the number of possible mixes of resources that may have contributed to any given residue’s formation will be exceedingly large. These and other independent variables mean that making inferences of paleodiet from stable carbon isotope analysis of charred cooking residues is not at this time practicable. Additional experimentation taking all or some of these variables into account is required to properly understand the relationships between food source d13C values and those obtained from charred cooking residues, and to make confident dietary inferences from those values. 8. Conclusions It is not possible to understand the relationships between agriculture and socio-economic developments without a firm understanding of crop histories. What has become increasingly clear over the past decade is the need to develop and use multiple lines of evidence to build those crop histories [46]. Reliance on single sources of evidence can result in erroneous conclusions about agricultural evolution and the roles played by crops in subsistence systems. Continued refinement of methods and techniques for identifying prehistoric food sources and their relative dietary contributions is critical. Morton and Schwarcz’s attempt to standardize approaches to understanding d13C values from prehistoric cooking residues was just such an attempt. However, the combined results of the three experiments and the modeling summarized here indicate that direct interpretation of d13C ratios to estimate the contribution of maize is not possible. In reference to the results of their isotopic analysis of prehistoric cooking residues, Morton and Schwarcz [31, p. 515] ask, ‘‘Why is there so little evidence of the presence of C4 plants. in these residues?’’ They suggest that maize was probably a minor ingredient in the foods cooked in pots and that rather, it must generally have been consumed as whole grains or on the cob [31, p. 515]. Our results suggest that the explanation probably lies in the nonlinear relationship between the proportion of maize cooked in a pot and the resultant d13C value of the residue. In order to interpret the d13C value, prior knowledge of the resources cooked in the pot is essential. Given this finding, it is unlikely that independent use of stable carbon isotope analysis of charred cooking residues is a viable technique for extracting paleodietary information. Acknowledgements Obtaining deer meat with low or no C4 content posed a serious problem to our experiments. We thank Dr. George Cornell, Director of the MSU Native American Institute, for putting us in touch with Kris and Dixie Stewart, proprietors of the Toonerville Riverboat Tour, Soo Junction, MI. The Stewarts provided us with appropriate samples of C3 consumer deer meat from Michigan’s Upper Peninsula. Funds for the deeremaize and wild riceemaize carbon isotope assays were made available by the New York State Museum 812 J.P. Hart et al. / Journal of Archaeological Science 34 (2007) 804e813 and the New York State Biodiversity Research Institute. Funding for the maizeechenopodium assays was from Geochron Research Awards and The Wenner Gren Foundation (Grant #6606). Dr. Mark Schurr, Dr. Robert Feranec, and Dr. Henry Schwarcz reviewed and provided many useful comments and suggestions on an earlier version of this paper. Dr. Stanley Ambrose provided many insightful comments on a version of the paper read by Lovis at the 2006 Society for American Archaeology meeting. His comments helped us to clarify a number of issues in this article. We appreciate our colleagues’ insights. We also thank two anonymous reviewers for their comments and suggestions. References [1] M.J. Adair, Great Plains paleoethnobotany, in: P.E. Minnis (Ed.), People and Plants in Ancient Eastern North America, Smithsonian Books, Washington, DC, 2003, pp. 258e346. [2] M. Bender, Mass spectrometric studies of carbon 13 variations in corn and other grasses, Radiocarbon 10 (1968) 468e472. [3] M. Bender, D.A. Barreis, R.L. Steventon, Further light on carbon isotopes and Hopewell agriculture, American Antiquity 46 (1981) 346e353. [4] N. Conard, D.L. Asch, N.B. Asch, D. Elmore, H. Gove, M. Rubin, J.A. Brown, M.D. Wiant, K.B. Farnsworth, T.G. Cook, Accelerator radiocarbon dating of evidence for prehistoric horticulture in Illinois, Nature 308 (1984) 443e446. [5] A.B. Cormie, H. Schwarcz, Stable isotopes of nitrogen and carbon of North American white-tailed deer and implications for paleodietary and other food-web studies, Palaeogeography, Palaeoclimatology, Palaeoecology 107 (1994) 227e241. [6] G.W. Crawford, D.G. Smith, V.E. Bowyer, Dating the entry of corn (Zea mays) into the lower Great Lakes, American Antiquity 62 (1997) 112e 119. [7] G.W. Crawford, D.G. Smith, J.R. Desloges, A.M. Davis, Floodplains and agricultural origins: a case study in south-central Ontario, Canada, Journal of Field Archaeology 25 (1998) 123e137. [8] J.W. Eerkens, GCeMS analysis of fatty acid ratios of archaeological potsherds from the Western Great Basin of North America, Archaeometry 47 (2005) 83e103. [9] R.P. Evershed, C. Heron, S. Charters, L.J. Goad, The survival of food residues: new methods of analysis, interpretation and application, in: A.M. Pollard (Ed.), New Developments in Archaeological Science. Proceedings of the British Academy 77, Oxford University Press, Oxford, 1992, pp. 187e208. [10] M.L. Fearn, K. Liu, Maize pollen of 3500 B.P. from southern Alabama, American Antiquity 60 (1995) 109e117. [11] G.J. Fritz, Prehistoric Ozark Agriculture: The University of Arkansas Rockshelter Collections, U.M.I. Dissertation Service, University of North Carolina at Chapel Hill, Ann Arbor, MI, 1986. [12] J.P. Gallagher, R.F. Boszhart, R.F. Sasso, K. Stevenson, Oneota ridged field agriculture in Southwestern Wisconsin, American Antiquity 50 (1985) 605e612. [13] J.P. Gallagher, R.F. Boszhart, R.F. Sasso, K. Stevenson, Floodplain agriculture in the driftless area: a reply to Overstreet, American Antiquity 52 (1987) 398e404. [14] K. Habfast, Fractionation in thermal ionization source, International Journal of Mass Spectrometry and Ion Processes 51 (1983) 165e189. [15] R.G. Harrison, M.A. Katzenberg, Paleodiet studies using stable carbon isotopes from bone apatite and collagen: examples from Southern Ontario and San Nicolas Island, California, Journal of Anthropological Archaeology 22 (2003) 227e244. [16] J.P. Hart, N.A. Sidell, Additional evidence for early cucurbit use in the northern Eastern Woodlands east of the Allegheny Front, American Antiquity 62 (1997) 523e537. [17] J.P. Hart, Maize agriculture evolution in the eastern woodlands of North America: a Darwinian perspective, Journal of Archaeological Method and Theory 6 (1999) 137e180. [18] J.P. Hart, D.L. Asch, C.M. Scarry, G.W. Crawford, The age of the common bean (Phaseolus vulgaris L.) in the Northern Eastern Woodlands of North America, Antiquity 76 (2002) 377e385. [19] J.P. Hart, R.G. Thompson, H.J. Brumbach, Phytolith evidence for early maize (Zea mays) in the northern Finger Lakes region of New York, American Antiquity 68 (2003) 619e640. [20] C. Hastorf, M.J. DeNiro, Reconstruction of prehistoric plant production and cooking practices by a new isotopic method, Nature 315 (1985) 489e491. [21] M.A. Katzenberg, Stable isotope analysis of archaeological faunal remains from Southern Ontario, Journal of Archaeological Science 16 (1989) 319e329. [22] M.A. Katzenberg, S.R. Saunders, S. Abonyi, Bone chemistry, food, and history: a case study from 19th century upper Canada, in: S.H. Ambrose, M.A. Katzenberg (Eds.), Biogeochemical Approaches to Paleodietary Analysis, Kluwer Academic/Plenum Publishers, New York, 2000, pp. 1e22. [23] M.A. Katzenberg, H.P. Schwarcz, M. Knyf, F.J. Melbye, Stable isotope evidence for maize horticulture and paleodiet in southern Ontario, Canada, American Antiquity 60 (1995) 335e350. [24] T.D. Knapp, Pits, plants, and place: recognizing late prehistoric subsistence and settlement diversity in the Upper Susquehanna drainage, in: J.P. Hart, C.B. Rieth (Eds.), Northeast Subsistence-Settlement Change A.D. 700e1300, New York State Museum Bulletin 496, University of the State of New York, Albany, 2002. [25] P.L. Koch, D.L. Phillips, Incorporating concentration dependence in stable isotope mixing models: a reply to Robbins Hilderband and Farley, Oecologia 133 (2002) 14e18. [26] M.J. Kohn, T.E. Cerling, Stable isotope compositions of biological apatite, Reviews in Mineralogy and Geochemistry 48 (2002) 455e488. [27] W.A. Lovis, Curatorial considerations for systematic research collections: AMS dating of a curated ceramic assemblage, American Antiquity 55 (1990) 382e387. [28] M.E. Malainey, R. Przybylski, B.L. Sherriff, The fatty acid composition of native food plants and animals of western Canada, Journal of Archaeological Science 26 (1999) 83e94. [29] B.D. Marino, M.J. DeNiro, Isotopic analysis of archaeobotanicals to reconstruct past climates: effects of activities associated with food preparation on carbon, hydrogen and oxygen isotope ratios of plant cellulose, Journal of Archaeological Science 14 (1987) 537e548. [30] G.W. Monaghan, W.A. Lovis, K.C. Egan-Bruhy, Earliest cucurbita in the Great Lakes region USA, Quaternary Research: An Interdisciplinary Journal 65 (2006) 216e222. [31] J.D. Morton, H. Schwarcz, Paleodietary implications from stable isotopic analysis of residues on prehistoric Ontario ceramics, Journal of Archaeological Science 31 (2004) 503e517. [32] A.E. Needham, The Uniqueness of Biological Materials, Pergamon Press, Oxford, 1965. [33] D.M. O’Brien, D.P. Schrag, C.M. del Rio, Allocation to reproduction in a hawkmoth: a quantitative analysis using stable carbon isotopes, Ecology 81 (2000) 2822e2831. [34] B.E. Perkle, Cucurbita pepo from King Coulee, southeastern Minnesota, American Antiquity 63 (1998) 279e288. [35] D.L. Phillips, P.L. Koch, Incorporating concentration dependence in stable isotope models, Oecologia 130 (2002) 114e125. [36] C. Pierce, Reverse engineering the ceramic cooking pot: cost and performance properties of plain and textured vessels, Journal of Archaeological Method and Theory 12 (2005) 117e157. [37] I. Poole, F. Braadbaart, J.J. Boon, P.F. van Bergen, Stable carbon isotope changes during artificial charring of propagules, Organic Geochemistry 33 (2002) 1675e1681. [38] J.M. Quigg, M.E. Malainey, R. Przybylski, G. Monks, No bones about it: using lipid analysis of burned rock and groundstone residues to examine Late Archaic subsistence practices in South Texas, Plains Anthropologist 46 (2001) 283e303. J.P. Hart et al. / Journal of Archaeological Science 34 (2007) 804e813 [39] E.A. Reber, Maize detection in absorbed pottery residues: development and archaeological application, Department of Anthropology, Harvard University, Cambridge, MA, 2001. [40] E.A. Reber, S.N. Dudd, N.J. van der Merwe, R.P. Evershed, Direct detection of maize in pottery residues via compound specific stable carbon isotope analysis, Antiquity 78 (2004) 682e691. [41] E.A. Reber, R.P. Evershed, How did Mississippians prepare maize? The application of compound-specific carbon isotope analysis to absorbed pottery residues from several Mississippi Valley sites, Archaeometry 46 (2004) 19e33. [42] R.F. Sasso, La Crosse region Oneota adaptations: changing late prehistoric subsistence and settlement patterns in the Upper Mississippi Valley, Wisconsin Archaeologist 74 (1993) 324e369. [43] M.J. Schoeninger, Reconstructing prehistoric human diet, in: T.D. Price (Ed.), The Chemistry of Prehistoric Human Bone, Cambridge University Press, Cambridge, 1989, pp. 38e67. [44] J.K. Schulenberg, The Point Peninsula to Owasco transition in central New York, Department of Anthropology, The Pennsylvania State University, University Park, PA, 2002. [45] H.P. Schwarcz, J. Melbye, M.A. Katzenberg, M. Knyf, Stable isotopes in human skeletons of southern Ontario: reconstructing paleodiet, Journal of Archaeological Science 12 (1985) 187e206. [46] J. Smalley, M. Blake, Stalk sugar and the domestication of maize, Current Anthropology 44 (2003) 675e703. 813 [47] R.G. Thompson, R.A. Kluth, D.W. Kluth, Tracing the use of Brainerd Ware through opal phytolith analysis of food residues, Minnesota Archaeologist 53 (1994) 86e95. [48] R.G. Thompson, J.P. Hart, H.J. Brumbach, R. Lusteck, Phytolith evidence for twentieth-century B.P. maize in northern Iroquoia, Northeast Anthropology 68 (2004) 25e40. [49] L.L. Tieszen, T.W. Boutton, K.G. Tesdahl, N.A. Slade, Fractionation and turnover of stable carbon isotopes in animal tissues: implications for d13c analysis of diet, Oecologia (Berlin) 57 (1983) 32e37. [50] L.L. Tieszen, T. Fagre, Carbon isotopic variability in modern and archaeological maize, Journal of Archaeological Science 20 (1993) 25e40. [51] C.S.M. Turney, D. Wheeler, A.R. Chivas, Carbon isotope fractionation in wood during carbonation, Geochemica et Cosmochimica Acta 70 (2006) 960e964. [52] N.J. van der Merwe, Carbon isotopes, photosynthesis, and archaeology, American Scientist 70 (1982) 596e606. [53] G.E. Wagner, Uses of plants by the Fort Ancient Indians, U.M.I. Dissertation Services, Washington University, Ann Arbor, MI, 1987. [54] C.D. White, H.P. Schwarcz, Ancient Maya diet at Lamanai, Belize: as inferred from isotopic and chemical analysis of human bone, Journal of Archaeological Science 16 (1989) 451e474. [55] D.R. Whitehead, Prehistoric maize in southeastern Virginia, Science 150 (1965) 881e883.
© Copyright 2026 Paperzz