Rurrent Cson, Arizona 85712-1201 USA ISSN: 0033-8222 .Our°Na1 RADIOCARBON an International Journal of Cosmogenic Isotope Research

VOLUME 41 / NUMBER 2 / 1999 Radiocarbon

An International Journal of Cosmogenic Isotope Research

Editor AUSTIN LONG

Consulting Editor AJ TJULL

Managing Editor Emerita RENEE S KRA

Managing Editor DAVID R SEWELL

Associate Managing Editor KIMBERLEY TANNER ELLIOTT C ''98 apartment of Geosciences D3 e University of Arizona 148 17 East Ft. Lowell Road rurrent cson, Arizona 85712-1201 USA ISSN: 0033-8222 .our°na1 RADIOCARBON An International Journal of Cosmogenic Isotope Research

Editor: AUSTIN LONG Consulting Editor: A J T JULL Managing Editor Emerita: RENEE S KRA Managing Editor: DAVID R SEWELL Associate Managing Editor: KIMBERLEY TANNER ELLIOTT Published by Department of Geosciences The University of Arizona

Published three times a year at The University of Arizona, Tucson, AZ 85712-1201 USA.

Subscription rate (1999): $120.00 (for institutions), $65.00 (for individuals). Foreign postage is extra. A complete price list, including Proceedings of International Conferences, special publications and back issues, appears on the inside back cover of this issue. Advertising rates available on request, or see http://www.radiocarbon. org/adrates.htm1.

Missing issues will be replaced without charge only if claim is made within three months (six months for India, New Zealand and Australia) after the publication date. Claims for missing issues will not be honored if non- delivery results from failure by the subscriber to notify the Journal of an address change.

Authors: See our "Information for Authors" document at http: / /www. radiocarbon. org/Authors / for guidelines concerning manuscript submission and format. All correspondence and manuscripts should be addressed to the Managing Editor, RADIOCARBON, Department of Geosciences, The University of Arizona, 4717 East Ft. Lowell Road, Tucson, AZ 85712-1201 USA. Tel.: +1 520 881-0857; Fax: +1 520 881-0554; Internet: editor@radiocarbon. org

List of laboratories. Our comprehensive list of laboratories is published annually, and is also available on the WWW at http: / /www. radiocarbon. org/ Inf o/ lablist. html. We ask all laboratory directors to provide their laboratory code designation, as well as current telephone and fax numbers, and e-mail addresses. Changes in names or addresses, additions or deletions should be reported to the Managing Editor. Convention- al and AMS laboratories are now arranged in alphabetical order by country and we include laboratories listed by code designation.

RADIOCARBON on the World Wide Web: ht tp : / /www.. radiocarbon. org/

RADIOCARBON is indexed and/or abstracted by the following sources: Anthropological Index; Anthropological Literature; Art and Archaeology Technical Abstracts; Bibliography and Index of Geology (GeoRef); British Archaeological Bibliography; Chemical Abstracts; Chemistry Citation Index; Current Advances in Ecological and Environmental Sciences; Current Contents (ISI); FRANCIS (Institut de l'Information Scientifique et Technique - CNRS); Geographical Abstracts; Geological Abstracts; Oceanographic Literature Review; Science Citation Index; Social Sciences Citation Index. Vol 41, Nr 2 Radiocarbon 1999

CONTENTS

FROM THE EDITOR Austin Long ......

ARTICLES

Methods, Materials, and Instruments Reliability of Bone Gelatin AMS Dating: Rattus exulans and Marine Shell Radiocarbon Dates from Pauatahanui Midden Sites in Wellington, New Zealand Nancy Beavan Athfield, Bruce McFadgen, Rodger Sparks ...... 119 Radiocarbon Dating of "Old" Charcoal Using a Wet Oxidation, Stepped-Combustion Procedure MI Bird, L KAyliffe, L K Fifield, CS M Turney, R G Cresswell, T T Barrows, B David ..... 127 Changes in 14C Activity over Time during Vacuum Distillation of Carbon from Rock Pore Water G R Davidson,I C Yang ...... 141 Radiocarbon Age Anomalies in Land Snail Shells from Texas: Ontogenetic, Individual, and Geographic Patterns of Variation Glenn A Goodfriend, G Lain Ellis, L J Toolin ...... 149 Archaeological and Historical Studies Change of Diet of the Greenland Vikings Determined from Stable Carbon Isotope Analysis and 14C Dating of Their Bones Jette Arneborg, Jan Heinemeier, Niels Lynnerup, Henrik L Nielsen, Niels Rud, Arny E Sveinbjornsdottir ...... 157 Use of Radiocarbon Dating in Assessing Christian Connections to the Dead Sea Scrolls GA Rodley, B E Thiering ...... 169 AMS 14C Dating of Equipment from the Iceman and of Spruce Logs from the Prehistoric Salt Mines of Hallstatt Werner Rom, Robin Golser, Walter Kutschera, Alfred Priller, Peter Steier, Eva M Wild ...... 183 DATE LIST Rudjer Boskovic Institute Radiocarbon Measurements XIV Nada Horvatincic, Bogomil Obelic, Ines Krajcar Bronic, Dugan Srdoc, Romana Calic .... 199 NOTES AND COMMENTS Radiocarbon Calibration by the Date Distribution Method Paul Muzikar...... 215

LETTER TO THE EDITOR Bias, Accuracy, and Precision Marian Scott ...... 221

RADIOCARBON UPDATES ...... 223

CORRECTION ...... 225

1 EDITORIAL BOARD

EDOUARD BARD Aix-en-Provence, France OWEN K DAVIS Tucson, Arizona, USA ELLEN R M DRUFFEL Irvine, California, USA CALVIN J HEUSSER Tuxedo, New York, USA SHEELA KUSUMGAR Ahmedabad, India STEVEN W LEAVITT Tucson, Arizona, USA ANN P McNICHOL Woods Hole, Massachusetts, USA ANDREW M T MOORE New Haven, Connecticut, USA PAVEL POVINEC Bratislava, Slovakia Monaco

MICHAEL B SCHIFFER Tucson, Arizona, USA E MARIAN SCOTT Glasgow, Scotland RODGER SPARKS Lower Hutt, New Zealand JOHANNES VAN DER PLICHT Groningen, The Netherlands

JOHN S VOGEL Livermore, California, USA WEIJIAN ZHOU Xi'an, China

11 FROM THE EDITOR

As this is my last comment here as RADIOCARBON'S editor, I take this opportunity to reflect on the journal's history, and where it may be headed in the next decade. As of the first of July this year, my title changed from Professor to Emeritus Professor. This is an appropriate juncture to consider which sectors of my professional life will be emphasized and which will be shifted to other persons. The journal editorship is one sector that will be turned over to others. Because this journal is part of the Geosciences Department at the University of Arizona, it was the responsibility of the Geosciences department head to appoint a new editor. Fortunately for the Journal, Tim Jull, who has worked closely with me on editorial matters for the past few years, has agreed to take over the editorship. He has been on the journal editorial staff as consulting editor since 1994, and has been involved with 14C measurements since the AMS facility first came to Arizona in 1981. Warren Beck and George Burr will back him up as associate editors. Both Warren and "Burr" are researchers in the University of Arizona AMS facility and have actively participated in recent Radiocarbon conferences. Warren is probably best known to you as one of the guest editors (with Ellen Druffel and Ann McNichol) of the "14C tribute to Reidar Nydal, Cycling and the Oceans". He has been an active participant in expan- sion of the radiocarbon calibration using corals. Burr worked with Meyer Rubin in the United States Geological Survey (Reston, Virginia) radiocarbon laboratory for 10 years before moving to Tucson in 1990. His current work includes 14C in groundwater and high-precision AMS 14C measurements in tree rings and fossil corals. With the insightful and energetic participation of these three active researchers in cosmogenic isotopes, I am certain the journal will be guided well in the future.

It seems that I have been only a few years in the editor capacity, but it has in fact been a decade. The present masthead makeover is only the most recent in a series of mostly evolutionary changes in editorship. RADIOCARBON had its origins as a supplement to the American Journal of Science. It originally was a repository of virtually all radiocarbon dates produced throughout the world and has steadily evolved toward publishing primarily research articles that illustrate applications of cosmogenic isotopes in earth science. Volume 1 appeared in 1959 with Richard Foster Flint and Edward S Deevey Jr as editors. In 1961 "supplement" was dropped from the title, replaced by the cover phrase "published annually by the American Journal of Science". Irving Rouse was added as a third editor in 1963. In 1968 J Gordon Ogden III was added as a fourth editor. In the same year Renee Kra became the managing editor. In 1972 Minze Stuiver replaced Deevey as one of the four editors, and in 1977.he became the senior editor. The 1978 volume showed Stuiver as senior editor and Ogden, Rouse, Stephen Porter, W M Mook and Hans Oeschger as associate editors. In 1984 Ronald B Davis replaced Ogden as associate editor. Annual volumes expanded from one issue per volume to two per volume in 1968. In 1973 RADIOCARBON produced three issues per year, the number that it has maintained ever since.

In 1989 the journal moved from Yale University in New Haven to the University of Arizona in Tuc- son with myself as editor. Renee Kra moved also to continue her role as managing editor. At this time the journal added a long list of rotating associate editors (henceforth to be known as the editorial board). In 1993 Dr David Sewell and Ms Kimberley Tanner Elliott became assistant editors. In 1994 Tim Jull was added as consulting editor. At present, David Sewell effectively handles the managing editorship in light of Renee Kra's recent illness.

Finally, I want to thank you for your past and present support of the journal and urge you to continue to submit your best research papers pertaining to the application of cosmogenic isotopes to chronological problems and understanding the behavior of natural systems. Austin Long

111 FROM THE MANAGING AND ASSOCIATE MANAGING EDITORS

Austin Long is one of those people who is notoriously averse to having a fuss made over him, but we did not want to let the occasion of his retirement pass without a few quiet words of appreciation.

As an editor, Austin has been unwavering in his devotion to concision, clarity, and exact language. We have learned from him not to tolerate calling a bar graph without class intervals a "histogram", and to substitute the more accurate "uncertainty" for what has often been called "error" in the reporting of radiocarbon dates. Whenever he has returned a set of proofs to the RADIOCARBON office, we knew that we would find pencil marks through any flabby language that we had let slip through: the idle "there are", the pretentious "it is to be expected that", and their kin.

He has been unwavering also in his devotion to objectivity and proper scientific method. The most difficult part of a scientific journal editor's job is to adjudicate the occasional paper where authors and reviewers violently disagree or accuse one another of bias. More than once Austin has put in long hours soliciting and reading second, third, or even fourth reviews of disputed papers, along with authors' revisions and explanations, before finally judging a submission to be publishable or not. And although we may have heard him grumble in private about the lunacy of a particular author or reviewer, he never allowed himself to reject a paper because of personal objections to it. The radiocarbon community is a highly diverse and sometimes contentious one; Austin's ability to remain calm and fair-minded has helped prevent its infrequent disputes from becoming serious. We thank him for bringing order out of chaos for this past decade, in the outside world as well as in our journal office.

David Sewell Kimberley Elliott Managing Editor Associate Managing Editor

RELIABILITY OF BONE GELATIN AMS DATING: RATTUS EXULANS AND MARINE SHELL RADIOCARBON DATES FROM PAUATAHANUI MIDDEN SITES IN WELLINGTON, NEW ZEALAND

Nancy Beavan Athfieldl Bruce McFadgen2 Rodger Sparks1

ABSTRACT. A suite of 6 bone gelatin accelerator mass spectrometry (AMS) radiocarbon dates for Rattus exulans Peale and 14C associated beta decay dates for Austrovenus stutchburyi shell are presented for 4 middens at Pauatahanui, Wellington, New Zealand. Mean calibrated age ranges of Rattus exulans (520-435 BP and 350-330 BP at 95% confidence level) and shell (465-375 BP at 95% confidence level) from the 4 midden sites overlap. The agreement between Rattus exulans bone gelatin dates and associated shell provides an inter-sample comparison of 14C dating using both gas counting (beta decay) and AMS dating techniques. We examine the adequacy of the standard gelatinization treatment for bone samples, which has been employed consistently at the laboratory since 1995.

INTRODUCTION

Bones of some Rattus exulans found in naturally deposited, avian predator middens in New Zealand have conventional radiocarbon ages up to 2 ka BP (Holdaway 1996). These ages are considerably older than the accepted timing for the human colonization of New Zealand of less than around 700 14C BP, based on analyses of dates for human occupation sites (Anderson 1991; McFadgen et al. 1994; Higham and Hogg 1997). Rattus exulans is not endemic to New Zealand. Its arrival on Pacific islands is associated only with transport by humans (Matissoo-Smith 1994) and 14C ages on subfossil specimens of Rattus exulans are thus a proxy for the times of earliest human contact with New Zealand.

An early date for the colonization of New Zealand by Rattus exulans has implications for understanding changes in New Zealand's unique biodiversity following human contact. It is therefore important to determine whether the date of about 2 ka BP for the arrival of Rattus exulans is reliable.

There has been some controversy as to the reliability of the earliest dates of 1500-2000 BP for Rat- tus exulans from avian predator sites. It has been suggested that the anomalous Rattus exulans ages are a dietary effect arising from rodents eating marine-based foods depleted in 14C (Anderson 1996). Beavan and Sparks (1998) have reviewed modern, island-based populations of Rattus exulans and observed small variations of 14C that could be associated with diet. The partitioning of diet between marine and terrestrial sources of food, however, appears to moderate the full effect of a marine depletion, and depletions seen in modern populations of Rattus exulans suggest that the effect on 14C ages would be <30014C yr. This is considerably less than required to explain the 800-1200-yr shift from approximately 700 BP in the oldest Rattus exulans dates from the natural sites, as in Hold- away's (1996) corpus of samples.

Anderson (1996) has also criticized the efficacy of the preparation methods used in obtaining Rattus exulans ages. At one occupation site, Pleasant River in North Otago, New Zealand, results from 2 laboratories, including Rafter, have Rattus exulans with calibrated ages up to 900 yr greater than the 400-800 cal BP time span of midden layers as based on 14C dates of associated shells and charcoals (Smith and Anderson 1998). The Rafter Lab examined soils from the site for significantly depleted carbon that could exchange with bone, yet found no obvious source of a tenacious contaminant. Work on the problem at this cultural site at Pleasant River will be described in a separate publication

1 Rafter Radiocarbon Laboratory, Institute of Geological and Nuclear Sciences, Lower Hutt, New Zealand 2New Zealand Department of Conservation, Science and Research Unit, PO Box 10420, Wellington, New Zealand

119 120 NB Athfield et al.

(Beavan and Smith forthcoming). In addition, a review of the reliability of bone gelatin dates from CaCO3-rich deposits in predator-deposited sites has been completed (Holdaway and Beavan 1999). In this paper we report on other tests of the efficacy of chemical preparation techniques employed at the Rafter Laboratory and used for all rat bone since 1995. We have used Rattus exulans mandibles from middens of known age on the southern shore of the Pauatahanui Harbor, north of Wellington, 14C shells from the same New Zealand (Figures 1 and 2). We compare the results to dates on marine middens. We also utilize a protocol for detecting the exchange of carbon between bones and their burial environment via amino acid and isotopic analysis of both bones and burial soils.

Porirua Harbour

4 km

pnuifigl

Figure 1 Location of Pauatahanui Inlet

Pauatahanui Middens Pauatahanui Harbour is an inland arm of the sea that formed when a river valley was drowned by the rise in sea level at the end of the last Pleistocene glaciation (Figure 1). Hills around the harbor are greywacke sandstone mantled by non-calcareous loess deposited during the Pleistocene glaciations (Healy 1980). Shell middens on the shoreline and hills around the harbor mark the sites of old cultivation grounds and temporary hunting and gathering camps used by the Maori in prehistoric times.

The 4 shell middens (Ml, M2, M4, site R27145; M9, site R27135) on hills behind the southeastern shore of the harbor (Figure 2) were excavated in 1978-1979. The middens were discrete deposits of shells originally discarded on the ground surface, covering an area between approximately 60 m2 and 225 m2. The maximum thicknesses of the deposits varied between the 4 middens from 10 to 35 cm. Over time, the shells gradually became incorporated into the topsoil. At the midden edges, where shell layers were thin, the shells had been scattered through the topsoil by soil biological activity. On the thicker parts of the middens, windblown drift had buried the shells and gradually Reliability of Bone Gelatin AMS Dating 121 percolated down through them. Soil-forming processes had not entirely destroyed the stratigraphy, and a few shell lenses comprising concentrations of burnt shell and various-sized shell could be dis- cerned. Apart from the few shell lenses, there were no other stratigraphic breaks.

All 4 middens contained the bones of Rattus exulans that had been hunted and trapped in the sur- rounding forest and grassland. The bones were stained and blackened by contact with the burial soil. Soils formed of greywacke and loess are typically moderately acid (NZ Soil Bureau 1968), which would contribute to the weathering of the bone. The preservation of the Rattus exulans bones within the middens was aided by the calcareous nature of the associated shell.

The shells from the middens (Ml, M2, M4, M9, Figure 2) were 14C-dated by the gas-counting (beta decay) method in 1979 (R A Empson and B G McFadgen, unpublished data). 14C dates for shells of the harbor bivalve Austrovenus stutchburyi (Ml, M2, M4, M9) and the ocean coast, rocky shore gas- tropod Haliotis iris (Ml) indicate that the middens were deposited between 375 and 465 cal BP (see Table 3). Archived samples of Rattus exulans jawbones from each of the 4 middens were submitted for 14C analysis by AMS in 1997.

0 200 m Contour interval 15 m Contour heights in meters above sea level

Figure 2 Locations of Pauatahanui shell middens with dated Rattus exulans bones

Laboratory Procedures

Rattus exulans Bone Samples Jaw bones were brushed to remove burial soil, washed and sonically cleaned in deionized water, and dried in a vacuum oven at 30 °C. Each sample was pulverized in mortar and pestle to <450 µm, and demineralized in 0.5M HCl while stirring at room temperature for 1 h. Collagen was filtered from the solution and gelatinized with 0.O1M HCl in a nitrogen atmosphere at 90 °C for 16 h. The gelatin was double-filtered through Whatman® GF/C and 0.45 µm Acrodisc® filters, and lyophilized to 122 NB Athfield et al. determine yields. An average of 4 mg of bone gelatin was combusted in sealed tubes with CuO and silver wire at 900 °C. The CO2 from the combustions was graphitized and analyzed by AMS after an aliquot had been taken for b13C analysis.

Shell Samples The shells were dated by conventional (gas counting) methods in 1979. X-ray diffraction showed between 0.65% and 0.25% calcite. Between 109 and 122 g of crushed shell was treated with 400 mL concentrated HC1. After drying, the evolved gas was purified by absorption in CaO at 500 °C and re- evolved by heating the resulting carbonate to 800-900 °C. The CO2 from each sample was stored for up to 3 weeks before counting to allow radon to decay.

Soil Samples

A 1.87 g sample of soil that was clinging to bones from Midden 4 (R27/45) was picked free of root hairs and shell fragments, sieved to <212 pm, and ground in a mortar and pestle. The soil was washed twice in 0.5M HCl at 85 °C for 30 min to remove carbonates, then filtered and rinsed to neu- trality. Residue and final wash solution were lyophilized, for a yield of 1.495 g. A 104.1 mg sample of this prepared soil was combusted as above yielding 4.0 mg of carbon, which was graphitized for AMS analysis of total organic carbon of the soil.

We then prepared an extract of free amino acids from the total organic carbon component of the soil with 5M HCI. The mixture was placed in a Pyrex® tube and sealed under nitrogen, then heated at 100 °C for 16 h. The extract was double-filtered through Whatman® GF/C and 0.45 µm Acrodisc® filters, and lyophilized to determine yields.

Amino Acid Analysis

Subsamples of 1 mg of hydrolyzed bone protein, and a sample of the amino extract from Midden 4 soil, were analyzed on a Waters Pico-Tag HPLC system by the Renal Research Unit of Wellington School of Medicine using the method of Negro et al. (1987).

Stable Isotope Analysis Carbon and nitrogen in the samples were analyzed by an ANCA-SL elemental analyzer in continuous-flow mode interfaced to a Europa Scientific Geo 20/20 mass spectrometer. Carbon and nitrogen isotopes were analyzed simultaneously from an average 1.5 mg of sample. The CO2 and nitrogen gases were resolved using chromatographic separation on a GC column at 85 °C, and analyzed for S13C and b15N and C/N ratios. Machine error values are ±0.1%o for carbon and ±0.3% or better for nitrogen.

RESULTS

Rattus exulans Bone Condition The amino acid composition of bone protein was examined to detect protein degradation or intrusion of free amino acid from soils, in the manner of Stafford et al. (1988). The amino acid profile of the Pauatahanui bones is compared to a Sigma® rat tail Type 1 collagen and the average of 9 amino-acid profiles from modern populations of Rattus exulans (Table 1). The amino-acid profiles of the Pauatahanui subfossil bones are in good agreement with the modern rat protein standards, and indicate minimal diagenetic change to the former. There is also no evi- Reliability of Bone Gelatin AMS Dating 123 dence of exogenous amino acid contamination in the purified protein when the Pauatahanui bones are compared to the profile of free amino acids in the burial soil. The Pauatahanui bones have stable isotope values (Table 2) within the expected range for terrestrial- based diet (DeNiro and Epstein 1978,1981; DeNiro 1985). In addition, the C/N ratios for Pauataha- nui bone protein (Table 2) are within the 2.9-3.6 range for well-preserved bone (DeNiro 1985).

Table 1 Amino acid values (residues/1000) from amino-acid profiles for Pauatahanui Rattus exulans and soil of burial compared with Sigma® rat tail standard and modern R. exulans bones Sigma® rat Modern Rattus exulans Pauatahanui Rattus Amino acid standard' averageb average Hydroxyproline 74.0 83.1 ± 3.4 3.5 Aspartic 47.8 47.0 ± 2.4 6.0 Glutamine 76.5 81.0 ± 3.4 6.6 Proline 112.4 111.0 ± 2.5 4.0 Glycine 300.0 310.0±8.2 13.4 Alanine 117.4 107.0±2.2 Arginine 49.1 49.9±0.6 ±0.3 aSigma® Rattus Type 1 collagen standard, Lot 36H3873 bAverage of 9 modern Rattus exulans from Kapiti Island, Wellington, New Zealand. From Beavan and Sparks (1998). Average of 6 Pauatahanui Rattus exulans samples, R21565/1,2,4,7,8,9

Table 2 C/N ratios and stable isotope data for Rattus exulans bones from Pauatahanui Archaeological b15N 13C Sample ID site nr (%) R21565/1 R27/45 (M4) 8.3 R21565/2 R27/45 (Ml) 8.1 R21565/4 R27/35 (M9) 5.5 R21565/7 R27/35 (M9) 6.9 R21565/8 R27/45 (M2) 9.9 R21565/9 R27/35 (M9) 9.3

Comparison of Rat Dates and Shell Dates

The ages of the rat bones are compared to the shell ages in Table 3. There is no a priori reason to expect the shell middens to all be contemporary; nevertheless the shell dates are not significantly different (T' (Ward and Wilson 1978) = 4.7, T'0.95, df=5 = 7.81). Because the shells are marine organ- isms and rats are terrestrial, their 14C ages are not directly comparable. However, if the rats are contemporary with the middens, the rat 14C ages should be a homogenous set and the calibrated mean ages of the shells and bones should overlap. Comparison of the rat dates shows that they are not significantly different (T' = 6.7, T'0.95, df=3 =11.07). Shells are corrected for the local marine reservoir effect by applying a aR correction of -30 ± 15 (McFadgen and Manning 1990). Calibration of the weighted means of each date set shows that their 95% confidence limits age ranges overlap by 30 yr.

Soil Organic Carbon

We examined the possibility that a soluble component of soil organic carbon, either humic acids or free amino acids from soil humics, could be incorporated into the bone. Free amino acids are formed by the degradation of soil organic matter, and are incorporated in mobile soil humic acids. If the soil 124 N B Athfield et al.

Table 3 Radiocarbon ages for rat bones and marine shells from Pauatahanui archaeological sitesa Rattus exulans Austrovenus stutchburyi with marine- Archaeological site nr CRA Lab nr calibrated age range nr R27/45 (Ml) 422± 84 NZA7932 ±67 630-415 BP R27/45 (M2) 452 ± 69 NZA7411 28 450-310 BP R27/45 (M4) 361 ± 68 NZA7044 33 510-395 BP R27/35(M9) 439 ± 71 NZA7410 R27/35 (M9) 548 ± 68 NZA7933 28 R27/35 (M9) 313 ± 68 NZA7934 BP Weighted means of CRAs 422 ± 30 T'=6.7,df=5 752±17 T'=4.7,df=3 Calibrated age range of 520-435 weighted means 350-330 465-375 BP (95%) BP (95%) aMarine 14C calibrations according to Stuiver and Braziunas (1993), with AR= 30 ± 15 (McFadgen and Manning 1990). Weighted means of conventional radiocarbon ages (CRA) and T' statistic according to Ward and Wilson (1978) (T' 0.95. df 11.07, T' 7.81.) Terrestrial 14C decadal calibrations according Stuiver and Pearson (1993) with no hemi- = 5 = 0.95, df = 3 = sphere offset. amino acids were contaminating the Rattus exulans bone protein, we might then expect an alteration of bone amino acid profile or an age for the bones which was different from the shells. To test for effects of burial environment on bone dates, we analyzed prepared soil and the extracted soil amino acids from Midden 4 burial soil for stable isotopes, determined the 14C ages of the fractions, and made a soil-extract amino acid profile. The total organic carbon component of the burial soil from around Rattus exulans samples from Midden 4 (R27/45) had a b13C of -27%0 (Table 4). Two targets made from the total organic carbon component returned conventional 14C ages of 386 ± 69 BP and 491 ± 69 BP (Table 4).

We then carried out the extraction of free amino acids from the total organic carbon component. The yield of amino acid extract from the total organic carbon component was less than 1%, and when combusted, yielded approximately 16% carbon.

The age for the soil free amino acid extract (47 ± 67 yr, NZA-8357) is less than the conventional radiocarbon age (CRA) of the Rattus exulans bones from Midden 4 (361 ± 68 yr, NZA-7044), and the insoluble residue of the soil after the amino acid extraction (675 ± 70 yr, NZA-8356) is greater (Table 4). While the soil amino acid shows a significantly younger 14C age than the bone, other evidence indicates that exogenous amino acids are unlikely to be affecting data derived from rat bone protein. The amino acid analysis of the soil extract has a noncollagenous profile (Table 1) which, when compared with the collagenous amino acid profile for the Pauatahanui Rattus exulans bones, indicates that free amino acids from the soil have not been incorporated into the bone protein. The amino acid extract of the soil has a b13C of -24.0%0 (Table 4), and the insoluble residue of the soil remaining after the hot-water extraction has a b13C of -26.6%0 (Table 4). Rattus exulans samples from all 4 middens have b13C ranging from -19.3%o to -20.9%0 (Table 2), indicating that the gelatinization process used in preparation of the Rattus exulans bones has removed soil contaminants. Reliability of Bone Gelatin ANTS Dating 125

Table 4 Stable isotope and AMS data for Rattus exulans bones from Pauatahanui, Midden 4, compared with components of the soil of burial

Material b13C %c S15N %c (95%) nr Rattus exulans bone -20.2 68 R27145 Midden 4 Soil (total organic carbon) -27.0 No data 69 491 ± 69 Amino acid extract of soil -24.0 7.12 67 Insoluble soil residue, after amino -26.6 data 70 acid extractions

CONCLUSION

Bone has long been seen as difficult to date. From the Rafter Laboratory's experience of dating over 70 Rattus exulans specimens from New Zealand and elsewhere in the Pacific, and some hundreds of other bone samples, we find that the confidence in ages returned for this material can be improved by employing analytical techniques to test for degradation and contamination.

Comparison of AMS dates for Rattus exulans bone and shell from the Pauatahanui middens is a typical method of crossdating associated materials to establish confidence in 14C ages returned for a particular site or specific level in a site. Holdaway (1996), for example, used the method to test the reliability of dates from a number of sites in the North and South Islands, most notably the agreement of associated bird bone, charcoal, and moa eggshell AMS dates in sequences bounded by the Taupo Pumice at Hukanui in Hawkes Bay.

In the Pauatahanui Rattus exulans samples, a standard Rafter Laboratory procedure has returned accurate ages for bone from an archaeological site, with respect to site stratigraphy and 14C ages for associated shell. We have also employed a combination of amino acid profiling and stable isotope (15N, 13C) analysis to evaluate possible burial contamination effects and the quality of the bone sample. The results presented here support the efficacy of a standard treatment employed at the lab since 1995, with typical agreement for bone and stratigraphically associated materials processed at the Rafter Laboratory, including the corpus of Rattus exulans and associated bone, charcoal, and eggshell dates from avian predator sites, as produced for Holdaway (1996).

ACKNOWLEDGMENTS

We would like to thank Dr Nick Greenhill, Wellington School of Medicine, for work on the amino analysis, and Dr Christine Prior for discussions involving bone dating and amino acid analysis. This research was supported by the Institute of Geological and Nuclear Sciences, Ltd, Non-Specific Out- put Funding (NSOF).

REFERENCES

Anderson AJ. 1991. The chronology of colonisation in 40(2):601-13. New Zealand. Antiquity 65:767-95. Beavan NR, Smith IWG. Radiocarbon dates from Rattus Anderson AJ.1996. Was Rattus exulans in New Zealand exulans: Further analysis of the Pleasant River case. 2000 years ago? AMS radiocarbon ages from Shag Forthcoming. River Mouth. Archaeology in Oceania 31:178-84. DeNiro MJ. 1985. Postmortem preservation and alter- Beavan NR, Sparks RJ. 1998. Factors influencing 14C ation of in vivo bone collagen isotope ratios in relation ages of the Pacific rat Rattus exulans. Radiocarbon to paleodietary reconstruction. Nature 317:806-9. 126 N B Athfield et al.

of the DeNiro MJ, Epstein S. 1978. Influence of diet on the dis- Polynesia. a novel approach: genetic analysis tribution of carbon isotopes in animals. Geochimica et Rattus exulans. Journal of the Polynesian Society Cosmochimica Acta 42:495-506. 103:75-87. Zealand. DeNiro MJ, Epstein S. 1981. Influence of diet on the dis- New Zealand Soil Bureau. 1968. Soils of New tribution of nitrogen isotopes in animals. Geochimica Part 3. NZ Soil Bureau Bulletin 26(3). of et Cosmochimica Acta 45:341-51. Negro A, Garbisa S, Gotte L, Spina M. 1987. The use Healy WB. 1980. Pauatahanui Inlet - an environmental reversed phase high performance liquid chromatogra- study. New Zealand Department of Scientific and In- phy and precolumn derivitization with dansyl chloride dustrial Research. DSIR Information Series 141. for quantification of specific amino acids in collagen Higham TFG, Hogg AG.1997. Evidence for late Polyne- and elastin. Analytical Biochemistry 160:39-46. sian colonization of New Zealand: University of Smith IWG, Anderson AJ.1998. Radiocarbon dates from Waikato radiocarbon measurements. Radiocarbon archaeological rat bone: the Pleasant River case. Ar- 39(2):149-92. chaeology in Oceania 33:88-91. Holdaway RN. 1996. The arrival of rats in New Zealand. Stafford TW, Brendel K, Duhamel RC. 1988. Radiocar- 13C 15N Nature 384:225-6. bon, and analysis of fossil bone: removal of Holdaway RN, Beavan, NR. 1999. Reliable 14C AMS humates with XAD-2 resin. Geochimica et Cosmo- dates on bird and Pacific rat Rattus exulans bone gel- chimica Acta 52:2257-67. atin from a CaCO3-rich deposit. Journal of the Royal Stuiver M, Braziunas TF. 1993. Modelling atmospheric 14C Society of New Zealand. Forthcoming. 14C influences and ages of marine samples to McFadgen BG, Knox FB, Cole TRL.1994. Radiocarbon 10,000 BC. Radiocarbon 35(1):137-89. calibration curve variations and their implications for Stuiver M, Pearson GW. 1993. High-precision bidecadal the interpretation of New Zealand prehistory. Radio- calibration of the radiocarbon time scale, AD 1950- carbon 36(2):221-36. 500 BC and 2500-6000 BC. Radiocarbon 35(1):1-24. McFadgen BG, Manning MR. 1990. Calibrating New Ward GK, Wilson SR. 1978. Procedures for comparing Zealand radiocarbon dates of marine shells. Radiocar- and combining radiocarbon age determinations: a cri- bon 32(2):229-32. tique. Archaeometry 20(1):19-31. Matissoo-Smith E. 1994. The human colonisation of RADIOCARBON, Vol 41, Nr 2,1999, p 127-140 ©1999 by the Arizona Board of Regents on behalf of the University of Arizona

RADIOCARBON DATING OF "OLD" CHARCOAL USING A WET OXIDATION, STEPPED-COMBUSTION PROCEDURE

M I Bird1 L K Ayliffe1'2 L K Fifield3 C S M Turney' R G Cresswell3 T T Barrows' B David4

ABSTRACT. We present results that validate a new wet oxidation, stepped-combustion procedure for dating "old" charcoal samples. An acid-base-wet oxidation (ABOX) pretreatment procedure has been developed that is used in place of the conventional acid-base-acid (ABA) pretreatment. Combustions and graphitizations are performed in a vacuum line that is insu- lated from the atmosphere by a second backing vacuum to eliminate the risk of atmospheric leakage into the line at any stage of the procedure. Combustions are performed at 3 temperatures (330 °C, 630 °C and 850 °C) with a graphite target produced from the CO2 evolved during each combustion step. In this way, the removal of any contamination can be monitored, and a high degree of confidence can be placed on the final age. The pretreatment, combustion, graphitization, and measurement blank for the procedure, based on the analysis of a "radiocarbon-dead" graphite, is 0.5 ± 0.5 tg C (16, n=14), equivalent to ± 0.04 0.02 pMC or an "age" of approximately 60 ka for a 1 mg graphite target. Analyses of a "radiocarbon-dead" natural charcoal after ABOX pretreatment and stepped combustion suggest that the total blank (including contamination not removed by pretreatment) may be higher than for graphite, ranging up to 0.10 ± 0.02 pMC. Additional experiments confirm good agreement with accepted values for the international low-14C "New Kauri" standard (0.16-0.25 pMC). They also confirm excellent reproducibility, with 3 separate dates on different aliquots of a charcoal sample from Ngarrabullgan Cave (Queen- 14C sland, Australia) ranging from 35.2 to 35.5 ka BP. It is also demonstrated that the ABOX pretreatment, in conjunction with the new vacuum line described here, is able to remove contamination not removed by the conventional ABA pretreatment, suggesting that the technique can be used to produce reliable 14C dates on charcoal up to at least 50 ka.

INTRODUCTION

Charcoal is a popular sample type for radiocarbon dating and can yield reliable dates from a large number of archaeological and geological environments. In most cases a simple acid-base-acid (ABA) pretreatment is all that is required to remove contaminants and provide reliable age estimates. However, in some cases it has been demonstrated that the ABA pretreatment does not remove all contaminants (e.g. Gillespie et al. 1992; Gillespie 1.997), and the problem of ensuring the complete removal of contaminants becomes increasingly severe as the age of the sample increases. Detecting sample contamination and verifying the reliability of the ages produced also become more difficult as the age of the sample increases. In practice this means that many laboratories will only 14C quote ages to about 40 ka BP (thousands of '4C years Before Present), with ages greater than this generally considered to be "infinite", or indistinguishable from procedural blanks.

The so-called "radiocarbon barrier" and the difficulty of ensuring that ages are reliable at <1 % modem carbon levels has limited research in many disciplines. One such problem surrounds the timing of human occupation of the Australian continent. Early 14C determinations suggested that humans first arrived about 40 ka BP (the "short" chronology summarized in Allen and Holdaway 1995; O'Connell and Allen 1998), whereas the advent of luminescence techniques in the 1980s suggested that humans may have arrived at 54-60 ka BP (the "long" chronology, Roberts et al. 1994). The large difference in ages between the short and the long chronologies is unlikely to be explained by any disparities resulting from the use of 2 independent dating techniques-14C and luminescence- representing 2 different time "clocks" (David et al. 1997). Proponents of the "long" chronology have suggested that the discrepancy is likely to be due to the influence of a small amount of contamina-

1Research School of Earth Sciences, Australian National University, Canberra A.C.T. 0200, Australia 2Present address: Laboratoire des Sciences du Climat et de l'Environnement, Unite Mixte de Recherche, CNRS-CEA, Domaine du CNRS, Batiment 12, Avenue de la Terrasse, 91198 Gif sur-Yvette Cedex, France 3Department of Nuclear Physics, Research School of Physical Sciences and Engineering, Australian National University, Canberra A.C.T. 0200, Australia 4Geography and Environmental Science, Monash University, Clayton, Victoria, 3168 Australia

127 128 MI Bird et al.

sample tion on the 14C dates, as contamination by 1 % of modern carbon will make a "14C-dead" have an apparent age of approximately 37 ka BP (Chappell et al. 1996). This explanation has been challenged by some archaeologists on several grounds, including the fact that although no archaeological 14C dates beyond 40 ka BP have been measured in Australia, a significant number of geological 14C dates beyond 40 ka BP are reported in the literature. It has been proposed that this would not be the case if contamination were a significant general problem (Allen and Holdaway 1995; O'Con- nell and Allen 1998). Resolution of this discrepancy is important for a refined understanding of early human migration and colonization of the continent as well as of the ecological effects accompanying human colonization. The production of reliable 14C dates on old archaeological charcoal samples can assist in this respect.

We believe that for a 14C date to be considered reliable at 50 ka, for example, the combined blank 14C from natural contamination and laboratory handling should be equivalent to a age of at least 55 ka. In addition, the ages must be reproducible within the stated error bounds, and the removal of all contamination must be verifiable. The results of this study show that these criteria can be met for charcoal, enabling the production of reliable 14C dates in the >40 ka BP range, using a wet oxidation pretreatment protocol to ensure removal of contaminants, and a stepped-combustion procedure to verify decontamination of the sample.

EXPERIMENTAL TECHNIQUES Samples The samples used for the verification of the procedure represent a range of natural and artificial charcoals and graphites, as listed in Table 1 and described below. They were chosen to provide information on the total blank for the procedure and to check that the ages produced are reproducible both internally and against previously generated results for well-characterized samples.

1. The Ceylon graphite is a natural geological "14C-dead" graphite that is used in this study to enable an estimate of the total procedural and measurement blanks for the wet oxidation, stepped-combustion technique. 2. A sample of "14C-dead" charcoal was selected to test the ability of the method to remove organic carbon contamination. The charcoal was obtained from a buried paleosol, the Coulter Pedoderm, exposed in a gully on the southwest flank of Black Mountain (Canberra, Australia). The unit is thought to have undergone pedogenesis during a period of climatic stability similar to the present, and is likely to be at least of last interglacial age. The unit is overlain by alluvial fan gravels stratigraphically equivalent to a valley-fill deposit in an adjacent valley dated by optically stimulated luminescence (OSL) to greater than 50 ka (T T Barrows and R G Roberts, unpublished data). The buried paleosol has been subject to infiltrating waters carrying organic compounds from the vegetated surface. Interactions between the charcoal and groundwater are evidenced by fine gypsum crystals deposited on the surface of some fragments. 3. An aliquot of the New Kauri standard (Hogg et al. 1995) was pyrolized in vacuo for this study 14C at approximately 500 °C to produce an artificial "charcoal" with a known low activity. Both pyrolized and "raw" samples of wood were used to test the pretreatment methodologies. 4. A crushed and homogenized sample of charcoal from Ngarrabullgan Cave (North Queensland, Australia) was prepared to assess the reproducibility of ages generated by the technique using charcoal from a previously dated unit. A substantial body of dates have been published for this unit, with 19 14C determinations spanning 28.8-34.614C ka BP (David et al. 1997). '4C Dating of "Old" Charcoal 129

Pretreatment

All pretreatment steps are carried out in either a laminated flow cabinet, a HEPA-filtered clean lab, or capped vials, with all vessels and implements precleaned either by combustion overnight at 550 °C (Pyrex®) or 850 °C (quartz), or by washing with a hot solution of O.1 M K2Cr2O7 in a 2M solution of H2SO4. All water used in the pretreatments is Milli-Q® grade.

Hand-picked fragments of charcoal are crushed in a mortar and pestle and a 10-200 mg aliquot is weighed into a 50 mL capped plastic Falcon® centrifuge tube. 20 mL of 6M HCl is added to the tube, which is capped and left to stand for 1 h, after which it is centrifuged and rinsed twice with Milli-Q® water. If the sample contains appreciable mineral impurities, a 1:1 mix of concentrated HCl and HF can be used in addition to a simple acid wash and the sample is left at 60 °C overnight to dissolve mineral phases, after which the solution is rinsed as above. 50 mL of 1M NaOH is then added to the tube, which is again left to stand for 30 min before rinsing once with Milli-Q® water. These steps are designed to remove acid and base soluble organic material, respectively.

Next, 30 mL of acid-dichromate oxidant solution (0.1M K2Cr2O7 in a 2M solution of H2SO4) is added to the tube, which is capped and heated to 60 ± 0.5 °C in a temperature-regulated hot box. After 14-24 h the tube is centrifuged, the supernatant discarded, and the remaining particulate material is washed twice with Milli-Q® water. Bird and Grocke (1997) have demonstrated that oxidation for 72 h under the above conditions will remove all organic carbon from the sample, leaving a residue of "oxidation resistant elemental carbon" (OREC). The proportion of OREC left after oxidation for 72 h depends on the nature of the charcoal and can range widely from 20 to 80%. The choice of 14-24 h is a compromise required to ensure that sufficient sample material is left for further analysis. With larger samples or denser charcoal fragments, the oxidation time can be increased to 72 h or longer.

After the final 2 rinses, the particulate material is transferred by Pasteur pipette to a capped 3 mL glass vial in Milli-Q® water. The vial is warmed on a hot plate at low temperature for 5 min, after which the supernatant, along with fine OREC and mineral impurities, is pipetted back into the centrifuge tube and a fresh aliquot of Milli-Q® water is added to the vial. The procedure is repeated until the supernatant above the sedimented particles is clear after the settling period. This step removes fine mineral impurities as well as fine OREC and also leaches some adsorbed dichromate from the surface of the larger particles. The fine residue can be kept for separate analysis if necessary. The cleaned particulate OREC is covered and allowed to dry on a hot plate for several hours, and is then weighed.

We call this pretreatment ABOX (acid-base-oxidation) to differentiate it from the conventional ABA pretreatment.

Combustion

Significant contamination of samples from atmospheric leakage into the extraction line can potentially occur during the combustion of the sample to C02, or during graphitization. The Pyrex and silica vacuum extraction line developed for these analyses eliminates the possibility of atmospheric leakage by backing all the valves in the line with a second vacuum that separates the interior of the line from the atmosphere (Figure 1). Thus all valves have a pumped volume behind the 0-ring through which the valve stem passes, and likewise, the sample loading and take-off points are protected by a vacuum between the inside of the line and the atmosphere. The vacuum-backed valves and metal "take-off" fittings are illustrated in Figure 2. Initial evacuation of the line is via a rotary pump, with final evacuation via a molecular drag pump backed by an oil-free diaphragm pump. pressure gauge

arut 1 I 1 1 1 { 1 1 11 f n i I r 1 I I I i i I1 1 1 1 1 11 1 I. l__ 1 '

us i I - fj 1 iii i/ ° I!?I a I I'n `II I I /I I /I I I F I I i o a I I vacuum-backed I I l l ! vacuum-backed { 1 . R R molecular I I C alon o fi ing I I I I Ca on fittin I o tt 1 g sieve tra i (see rigure [) i i Q ee riyuIc ) conflatflange I I' I to baratron I

! I I I I I I I I II 1

I I I I E I I I I I 'backin vacuum line'

i i i i I Y I Y \2' H 'main line' I 1 I I I I I 11 I 11 i

I (under) ! I 1

ip ! I I l l I I I I 1 F II I i I I I 1!! 11 i (2 Figure 1 The combustion-ragP hitization line used for this study. Vacuum-backed valves to 13), manufactured to specifications by Embell Scientific Pty Ltd, and vacuum-backed 0-ring fittings (D and F) are shown in detail on Figure 2. The left side of the line is used for combustion of the sample the right side for graphitizations, and the middle of the line is used for qquantification of CO 2 evolved and dosingg b the combustion and graphitization sides with 02 or H2 as required (using a 0-1000 torn Baratron® capacitance manometer at I). The line is evacuated at valves 2 3 and 12, with initial evacuation via a rotary pump (at A) and then via a molecular drag pump (at B). 0z and H2 is admitted via valve 13, and precleaned by sorption onto the molecular sieve at H. Monitoring of pressure is via thermal conductivity gauges at C, E, and G. 14C Dating of "Old" Charcoal 131

to line via glass/metal transition

in line

Figure 2 Diagrams of (1) a Pyrex® vacuum-backed valve, with inlets and outlets of 12 mm OD, and (2) Cajon® fittings welded together to provide a pumpable volume between the 0-ring seal to the vacuum line and the atmosphere. The fittings were manufactured using: (A) 1/4 inch to 3/8 inch Cajon adaptor, drilled through to allow the passage of the 6 mm OD silica combustion tube; (B) 1/4 inch Cajon fitting with a welded "hood" of 1 inch OD, with outlets to the main vacuum line and the backing vacuum line. The pumped volume (C) is created by sealing the hood with a 1 inch Cajon fitting (D) welded to (A). All seals are made by the compression of Viton® 0-rings at the locations indicated on the diagram.

Prior to loading a sample, 200 mg of combusted CuO (850 °C overnight) is weighed into a silica tube (6 mm OD x 20 cm length), and both the tube and CuO are again combusted. Approximately 5-15 mg of OREC is weighed into a small combusted silica capsule (3.4 mm OD x 10 mm length), which is pushed down the silica combustion tube to rest inverted on the CuO at the bottom (Figure 2). A plug of 0.05 mm diameter >99.9% Ag wire (Aldrich; combusted at 550 °C overnight) is pushed down the tube to rest on top of the sample. This prevents particles from being entrained into the vacuum line during pump-down and removes any halogens or sulfur produced during the combustion.

The sample tube is loaded onto the line and evacuated. Whenever the line is opened, it is back- flushed with H2 to prevent the adsorption of atmospheric CO2 onto internal surfaces. For the initial combustion step, 99.99% 02 (Linde) is first frozen onto a type 4 molecular sieve with liquid nitrogen (N2(1)). The N2(1) trap is removed and replaced by a dry ice-ethanol trap (approximately -80°C) that allows O2 to sublime but will retain any carbon-bearing gas species. The sublimation is allowed to continue until a p02 of about 0.5 atm (sufficient to oxidize about 1 mg of carbon) is present in the combustion volume. CuO does not provide available oxygen during the first combustion step at 330 °C. Excess O2 is pumped off and the molecular sieve is baked at about 200 °C until the normal background pressure is reached prior to its next use. 132 MI Bird et al.

A furnace at 330 °C is placed around the sample and the temperature maintained for a 2 h period. Work by Cachier et al. (1989) has shown that this treatment will remove organic contaminants with minimal removal of black carbon from aerosol samples. This step serves to remove contaminants that may have been introduced during the loading procedure or that are adsorbed onto the sample surface. In practice it was found that pretreated OREC samples were partly susceptible to oxidation during this step, probably due to catalysis of the oxidation reaction by residual chromium ions on the sample surface (Mull et a1.1998). At the end of the oxidation time, the furnace is lowered, CO2 is is measured using an frozen with N2(1), and excess 02 is pumped off. The amount of CO2 produced MKS-Baratron® capacitance manometer (0-1000 torn). This CO2 is then graphitized, as described below.

The combustion volume is then re-isolated and heated at 630° C for 1 h. At this temperature, CuO provides the oxygen for the oxidation. As previously, the CO2 produced during this combustion step is purified, measured, and graphitized. The sample is then heated at 850 °C for a further hour, with the CO2 again purified, quantified, and graphitized. This last combustion step is repeated to ensure complete combustion of the sample. The quantity of CO2 produced during this final step is usually negligible and is discarded.

A graphite target is generally made from the CO2 produced during each combustion step (i.e., 330 °C, 630 °C, and 850 °C). In this way it is expected that any contaminants remaining after the 14C sample pretreatment will be combusted in the 330 °C step, and that the reliability of the age can be assessed by the coherence of the results from the 630 °C and 850 °C steps. This is because oxidation can be expected to remove less resistant phases first, and work progressively from the outside to the interior of the particles. In the case of untreated graphite, very little CO2 was produced in the 330 °C or 630 °C combustion steps, so the 3 targets were made from 3 combustions at 850 °C (850 I, II, III; Table 1). Graphitization The graphitization procedure is based on the conventional technique of iron-catalyzed reduction of CO2 to graphite in the presence of H2 at 630 °C, with the water produced during the reaction being frozen into a dry ice-ethanol trap (Vogel et a1.1984; Kitagawa et al. 1993; Gagnon and Jones 1993). An additional step used in our procedure is a preoxidation of approximately 0.5 to 1.0 mg iron powder (Merck; 10 µm) to remove carbonaceous contaminants present in the Fe powder, or introduced during the sample loading procedure. Approximately 0.5 atm of 02 is sublimed from the molecular sieve trap into the graphitization tube containing the iron powder and this is then heated to 330 °C for 2 h. Prior to graphitization, the iron powder is re-reduced with approximately 0.5 atm of to affect 99.999% H2 (Linde) at 630 °C for 2 h. The precombustion of the Fe powder does not appear its catalytic action during the conversion of CO2 to graphite. As for the 02, any H2 used in the procedure is purified prior to use by means of the molecular sieve trap.

Sample CO2 is frozen into the graphitization tube, which is then dosed with a l.lx stoichiometric excess of H2. The graphitization is conducted at 630 °C for 16-18 h. In general, CO2 produced during the 330 °C (A) and 630 °C (B) steps is graphitized in flame-sealed silica tubes offline, while the 850 °C (C) graphitization is conducted attached to the extraction line. The completeness of the graphitization for the 850 °C (C) fraction can then be checked by measuring the total pressure, and CO2 pressure at the end of the reaction period. During the graphitization, water produced from the reduction of CO2 by H2 is frozen into a dry ice-ethanol trap placed on the bottom of the graphitization tube. The graphite that is produced is stored in vacuo until analysis by flame-sealing the graphitization tube above and below the small bucket containing the graphite and Fe powder. 14C Dating of "Old" Charcoal 133

In practice, the initial combustion steps as well as the preoxidation and reduction of the iron powder for graphitization occur in parallel during the day, while the graphitizations take place overnight. One sample, yielding 3 targets, can be processed per 24 h using this procedure.

Accelerator Mass Spectrometry

The mass of graphite plus iron powder is determined as a yield check and then the graphite is pressed into 1 mm diameter holes in aluminum sample holders. These are loaded, together with 3 or 4 ANU sucrose standards, into a 32-sample wheel for insertion into an NEC multi-sample negative ion source. The 14C/13C ratio of each sample is measured by accelerator mass spectrometry using the 14UD accelerator at the Australian National University. Although the 613C value has not been measured, it has been assumed that ANU sucrose and charcoal have values of -11%o and -25%, respectively, and hence a fractionation correction of 14% (equivalent to 110 yr) has been applied to each of the charcoal samples.

A blank correction of 0.07 ± 0.05 pMC has also been subtracted from the measured ratios of the New Kauri and Ngarrabullgan charcoal samples, but not the Ceylon graphite analyses or the analyses of the Black Mountain charcoal samples. This blank represents the total spread of values obtained for both the Ceylon graphite and Black Mountain charcoal (as discussed below).

RESULTS AND DISCUSSION

The results of all analyses are presented in Table 1. The analyses of Ceylon graphite (generally 3 targets per analysis) represent the combustion, graphitization, and measurement blank for the procedure. All 14 measurements are consistent with the mean value of 0.04 pMC (x2v =1.2) and the standard deviation of the data set is 0.02 pMC (Figure 3). This blank is at least an order of magnitude lower than the 0.7% modern carbon that would be equivalent to an age of 40 ka, and is 2-10 times

0 0.00 0.02 0.04 0.06 0.08 0.10

pMC

Figure 3 Histogram of blank values obtained from 14 combusted and graphitized Ceylon graphite samples 14C Table 1: pMC and ages for a range of natural and artificial charcoals and graphites measured in this study. Quoted mean 14C ages are based upon amount-weighted pMC values for the 3 combustion fractions, and are equivalent to the "total combustion" age of the sample. Sample Description Pretreatment No. wt age time (h) (mg) (16) BP) Ceylon Geological graphite, Untreated (850) 1 -2510 graphite 14C "dead" ANUA-8126b (850) 1 -2620 ANUA-8127b CIII (850) 3 -1650 ANUA-9300bc CIII (850) 3 -3020 Untreated ANUA-8208b Cl (850) 1 -2530 ANUA-8209b CII (850) 1 -4350 ANUA-8210b CIII (850) 1 -3310 Untreated ANUA-9424b CII(850) 1

Untreated ANUA-9422b Cl (850) 1 -2410 ANUA-9423b CII(850) 5

ABOX ANUA-9426b CI (850) 1 -2110

ABOX ANUA-9425b CIII (850) 1

ABOX ANUA-9427b Cl (850) 1 -3350 ABOX ANUA-9428b Cl (850) 1 95-BMtn- Natural charcoal from Untreated A (330) 2 -550 FAN-002 >50 ka sediments (Bar- ANUA-10322 (630) 1 rows, unpublished 220 / -1060 data) ANUA-10323 C(850) 1 -1440 Mean 0.56 0.03 -460 ABA ANUA-9415 A (330) 2 -390 ANUA-9416 B (630) 1 -990 ANUA-9417 C (850) 1 -1210 Mean 0.32 0.02 -540 ABOX ANUA-9418 A (330) 2 -680 ANUA-9419 B (630) 1 -1430 ANUA-9420 C (850) 1 -1660 Mean 0.17 0.02 -830 ANUA-6130 C (850) 4 -1750 New Kauri Kauri wood (0.12-0.21 Pyrolized ABOX (850) 3 -1830 standard pMC), Hogg et al. ABOX (850) 2 -2640 995 . Pyrolized ABOX ANUA-8802 (850) 1 -2040 ABOX ANUA-10308 C (850) 1 -2930 Sample Description Pretreatment No. wt age time (h) (mg) (ia) BP) MM25 Archaeological char- Untreated A (330) 2 -460

XU8 (3) coal, Ngarrabullgan ANUA-9308 (630) 1 -450 Cave, Queensland ANUA-8803 C (850) 1 -580 See David et al. (1997) Mean 1.47 -380 for site description HCI/NaOH ANUA-9309 (330) 2 -590

ANUA-9310 B (630) 1 /-660

ANUA-8804 C (850) 1 -560 Mean 1.44 0.08 HCI/HF/NaOH ANUA-9312 A (330) 2 -520

ANUA-9313 B (630) 1 -540

ANUA-8805 C (850) 1 -530 Mean 1.68 0.09 ABOX (7 h) ANUA-9314 A (330) 2 -450

ANUA-8806 C (850) 1 -690 ABOX (14 h) ANUA-9315 A (330) 2 -460 ANUA-9316 B (630) 1 -540 ANUA-8807 C (850) 1 -660 Mean 1.37 0.07 aWith the exception of the Ceylon 9raphites and the Black Mountain charcoal, a blank correction of 0.07 t 0.05 pMC has been subtracted from the measured PMC for each sample. bNe9li9ible gas at 340 °C and 650 °C. CRepeat measurement of ANUA-8127 after heating target in vacuo at 650 °C.

L) 136 MI Bird etal. lower than commonly reported procedural blanks (Kitagawa et al. 1993; Vogel et al. 1987; Gagnon and Jones 1993). Note that 4 of the samples, ANUA-9425 to ANUA-9428, were derived from Cey- lon graphite that had been subjected to the ABOX treatment before subsamples spanning a range of weights were taken for subsequent combustion and graphitization. The values obtained from these 4 samples are statistically indistinguishable from the 10 samples that were not subjected to any pretreatment prior to combustion.

The samples spanned a range of graphite weights from 0.1 to 2.2 mg, and the corresponding amount of modern carbon that would account for the observed pMC varies between about 0.1 and 1.7 µg, giving an average of approximately 0.5 ± 0.5 µg (16, n=14, equivalent to -60 ka for a 1 mg graphite target). It should also be noted that calculated blanks in excess of 1 µg were from the first 2 Ceylon graphites prepared using the combustion line, and calculated blanks for all subsequent analyses have been consistently lower. There is, however, no obvious correlation between sample weight and the measured 14C activity (Figure 4), suggesting that the few '4C atoms observed may already be present in the Ceylon graphite itself or were possibly added during target pressing.

0.10 o Untreated 0.09 ABOX 0.08

0.07

0.06

0.05

0.04

0.03

0.02

0.01

0.00 0.0 0.5 1.0 1.5 2.0 2.5

Sample size, mg Figure 4 Plot of the relationship between sample size and pMC values for untreated and ABOX Ceylon graphite targets. Error bars represent the 68% (±l6) range for overall analytical confidence.

Vogel et al. (1987) have suggested that diffusion of carbon through the walls of Vycor combustion tubes may occur during extended combustions at elevated temperature. To test this possibility for the silica tubes used in this study, samples of Ceylon graphite were combusted for 1 and 5 h (ANUA- 9422 and -9423, respectively; Table 1). The resulting 14C ages of 60.7 and 62.4 ka BP (0.05 and 0.04 pMC, respectively) are statistically indistinguishable and indicate that diffusion of modern carbon into the tubes during combustion does not contribute significantly to the 14C background. The Black Mountain charcoal sample (Figure 5) enables the estimation of the blank achievable for known "14C-dead" charcoal. This sample represents a worst case in that it is overlain by a vegetated 14C Dating of "Old" Charcoal 137

60000 ,

50000

Li 40000 4 A

a Untreated 30000 ABA ABOX

20000

330 °C 630 °C 850 °C Mean Fraction

Figure 5 Variations in untreated, ABA- and ABOX-treated 14C ages from Black Mountain charcoal at the 330 °C, 630 °C and 850 °C temperature fractions and amount-weighted mean ages for total combustion. Error bars represent the 68% (±16) range for overall analytical confidence. Note the consistently older ages derived using the ABOX-treatment. soil profile and has been exposed to infiltrating waters and the organic compounds carried by those waters since deposition of the sediments and subsequent pedogenesis. This is in contrast to charcoal from Ngarrabullgan Cave, for example, which has been substantially protected from interaction with infiltrating waters (discussed below). The Black Mountain charcoal is therefore an ideal sample with which to compare the relative efficacy of the ABOX and ABA pretreatments and stepped combustion. The results suggest 2 things;

1. Neither the stepped combustion of an unpretreated sample nor an ABA pretreatment with stepped combustion is sufficient to remove all contaminants. The 14C ages of the untreated (50.4 ka) and ABA (52.2 ka) 850 °C fractions are younger than the same fraction of the ABOX pretreated sample (55.6 ka). The weighted mean 14C age (equivalent to the age that would have been obtained from complete combustion) of the untreated and ABA pretreated samples for all combustion steps (41.6 and 46.3 ka) are significantly younger at the 26 level than the cone- sponding weighted mean age for the ABOX pretreated sample (51.3 ka). 2. The ABOX pretreatment alone does not remove all contamination, as the lower temperature fractions of the step-combusted ABOX pretreated sample have younger 14C ages (37.7 and 51.3 ka) than the 850 °C fraction (55.9 ka). It therefore seems that although total combustion can produce reliable ages on charcoal, the effective limit of the 14C technique will be considerably younger for samples that are analyzed by total combustion. The results also suggest that the blank for the ABOXlstepped-combustion procedure applied to charcoal from oxidizing surficial environments may be higher than for a crystalline graphite such as the Ceylon graphite. A conservative approach to assigning a blank value to analyses of natural charcoals 138 M I Bird et al. would be to assume that the true blank lies between the average value obtained for the ABOX pretreated 850 °C fraction of the Black Mountain charcoal and the blank obtained for the Ceylon graphite (i.e., 0.07 ± 0.05 pMC). This blank has been subtracted from the samples discussed below. Fur- ther analysis of known "14C-dead" charcoals from a variety of environments should allow a refinement of this value in future. The results for the 14C New Kauri standard indicate that the pretreatment procedure does not intro- duce contaminants. Regardless of whether the wood was pyrolized in vacuo or pretreated raw, the results for this standard, which range from 0.16 ± 0.06 pMC to 0.25 ± 0.06 pMC (48-51.7 ka), are internally consistent and comparable within error to the accepted range of values for this standard (0.12-0.21 pMC; Hogg et al. 1995). The series of measurements on subsamples of the homogenized charcoal from Ngarrabullgan Cave was designed to test various steps in the procedure for a typical archaeological sample. The 3 combustion steps for the 5 pretreatments returned 14C ages from 30.9 to 35.5 ka 14C BP (2.12-1.21 pMC, Table 1). For ABOX-pretreated samples, the results indicate that consistent, reproducible ages can be obtained from the 850 °C fraction, and possibly from the 630 °C fraction, but not from the 330 °C fraction (Figure 6). The results for the step-combusted, untreated sample further suggest that for relatively uncontaminated charcoals, the 850 °C fraction of the step combustion alone can produce a reliable age. For the samples that were subjected only to an acid-base pretreatment, it is evident that minor contamination is introduced to the sample during these initial steps, most likely from adsorption of atmospheric CO2 by the alkaline solution. However, this contamination is removed during the subsequent acid oxidation step of the full ABOX procedure. No ABA-treated sample was ana-

37000 ,

36000

35000

o Untreated m 34000 I HCI/NaOH (AB) V HCI/HF/NaOH (AB) 33000 ABOX 7 h ABOX 14 h 32000

31000

30000 330 °C 630 °C 850 °C Mean

Fraction

Figure 6 Variations in untreated, AB-, and ABOX-treated 14C ages from Ngarrabullgan charcoal at the 330 °C, 630 °C and 850 °C temperature fractions and weighted means. Error bars represent the 68% (±16) range for overall analytical confidence. Note the consistently older ages derived using the ABOX treatment. 14C Dating of "Old" Charcoal 139 lyzed, as the results from the untreated and ABOX-treated samples indicate that the Ngarrabullgan sample is virtually uncontaminated.

All 5 samples show a significant shift to older ages from the 330 °C to 630 °C step (Table 1 and Fig- ure 6). However, in the case of the AB-treated samples, there is a suggestion that the ages become younger for the 850 °C fraction, possibly owing to formation of Na2CO3, which has a decarbonation temperature of 851 °C. The results for the 850 °C fraction of the untreated sample and of the 2 ABOX-treated samples cluster tightly between 35.2 and 35.514C ka BP. This age is at the upper end of the range of 19 dates on charcoal from this level of the deposit (David et al. 1997), which range from 28.8 to 34.614C ka BR A sample previously analyzed from this deposit subjected to the ABOX pretreatment (without stepped combustion) also returned an age at the upper extreme of the dated sample population (OZC-733, 35.5 ± 0.6 ka; Roberts et al. 1994), again suggesting that the ABOX/ stepped-combustion procedure consistently removes contaminants not removed by conventional pretreatments.

CONCLUSION

This study has demonstrated that the ABOX pretreatment is a reliable technique for removing contaminants from charcoal and that the stepped-combustion procedure can ensure that blanks are suf- ficiently low to enable reliable 14C age determinations in the 40-50 ka time range. The production of multiple targets by stepped combustion provides an additional indicator of the reliability of age determinations in this time range. Previous studies have demonstrated that the major source of introduced contamination in the combustion of samples for 14C analysis comes from the copper oxide used during combustion (Vogel et al. 1987; Vandeputte et al. 1998). This suggests that the major contribution to reducing the overall blank for charcoal samples in this study comes from the use of stepped combustion, with oxygen employed as the oxidant at 330 °C. The stepped combustion removes contaminants present in the copper oxide and/or introduced during sample loading, and also removes contamination from samples not removed by the ABOX pretreatment. The results for the Black Mountain sample suggest that both the ABOX pretreatment and stepped combustion are required to reduce blanks to a level low enough to allow the measurement of reliable finite dates on charcoal in the 40-50 ka time range. The other procedures used in this study, including preoxidation of the iron powder, use of molecular sieves, storage of samples in vacuo and the use of a vacuum-backed extraction line, may also contribute to overall blank reduction, but more importantly, guard against random contamination of individual samples. Based on the range of blank values obtained from "14C-dead" charcoal and graphite, it seems likely that the appropriate blank for a natural charcoal sample will vary depending on the nature of the charcoal, its degree of alteration, and the environment from which it was obtained.

ACKNOWLEDGMENTS This work was supported by a grant from the Australian Institute of Aboriginal and Tones Strait Islander Studies. MIB and CT acknowledge the provision of fellowships from the Australian Research Council. We would like to thank S Robertson for the provision of an aliquot of the new Kauri standard, C Morgan for the manufacture of the "vacuum-backed" Cajon® 0-ring fittings, and Emmanuel Bellantoni of Embell Scientific Pty Ltd for the manufacture of the "vacuum-backed" glass valves. Chris Tomkins and Paul Middlestead assisted with the building and commissioning of the vacuum line. 140 M I Bird et al.

REFERENCES

Allen J, Holdaway S. 1995. The contamination of Pleis- background IAEA 14C quality assurance material. Ra- tocene radiocarbon determinations in Australia. An- diocarbon 37(2):797-803. tiquity 69:101-12. Kitagawa H, Masuzawa T, Makamura, T, Matsumoto E. Bird MI, Grocke DR. 1997. Determination of the abun- 1993. A batch preparation method for graphite targets dance and carbon isotope composition of elemental with low background for AMS 14C measurements. Ra- carbon in sediments. Geochimica et Cosmochimica diocarbon 35(2):295-300. Acta 61:3413-23. Mull G, Zhu WD, Kapteijn F, Moulijn JA. 1998. The ef- Cachier H, Bremond MP, Buat-Menard P. 1989. Determi- fect of NOX and CO on the rate of transition metal ox- nation of atmospheric soot carbon with a simple ther- ide catalysed black carbon oxidation - an exploratory mal method. Tellus 41B:379-90. study. Applied Catalysis B - Environmental 17:205- Chappell J, Head J, Magee, J. 1996. Beyond the radiocar- 20. bon limit in Australian archaeology and Quaternary O'Connell JF, Allen J. 1998. When did humans first ar- research. Antiquity 70:543-52. rive in greater Australia and why is it important to David B, Roberts R, Tuniz C, Jones R, Head J. 1997. know? Evolutionary Anthropology 6:132-46. New optical and radiocarbon dates from Ngarrabull- Roberts RG, Jones R, Spooner NA, Head MJ, Murray gan Cave, a Pleistocene archaeological site in Austra- AS, Smith MA. 1994. The human colonisation of Aus- lia: implications for the comparability of time clocks tralia: optical dates of 53,000 and 60,000 years bracket and for the human colonization of Australia. Antiquity human arrival at Deaf Adder Gorge, Northern Terri- 71:183-8. tory. Quaternary Science Reviews 13:575-83. Gagnon AR, Jones GA. 1993. AMS-graphite target pro- Vandeputte K, Moens L, Dams R. 1998. Study of the 14C- duction methods at the Woods Hole Oceanographic contamination potential of C-impurities in CuO and Institution during 1986-1991. Radiocarbon 35(2): Fe. Radiocarbon 40(1):103-10. 301-10. Vogel JS, Nelson DE, Southon JR. 1987.14C background Gillespie R. 1997. Burnt and unburnt carbon: dating levels in an accelerator mass spectrometry system. Ra- charcoal and burnt bone from the Willandra Lakes, diocarbon 29(3):323-33. Australia. Radiocarbon 39(3):225-36. Vogel JS, Southon JR, Nelson DE, Brown TA. 1984. Per- Gillespie R, Hammond AP, Goh KM, Tonkin PJ, Lowe formance of catalytically condensed carbon for use in DC, Sparks RJ, Wallace G. 1992. AMS radiocarbon accelerator mass spectrometry. In: Wolfli W, Polach dating of a Late Quaternary tephra site at Graham's HA, Anderson HH, editors. Proceedings of the 3rd In- Terrace, New Zealand. Radiocarbon 34(1):21-8. ternational Symposium on Accelerator Mass Spec- Hogg AG, Higham T, Robertson S, Beukens R, Kan- trometry. Nuclear Instruments and Methods in Phys- kainen T, McCormack FG, van der Plicht J, Stuiver ics Research B233:289-93. M. 1995. Radiocarbon age assessment of a new, near- RADIOCARBON, Vol 41, Nr 2, 1999, p 141-148 ©1999 by the Arizona Board of Regents on behalf of the University of Arizona

14C CHANGES IN ACTIVITY OVER TIME DURING VACUUM DISTILLATION OF CARBON FROM ROCK PORE WATER

G R Davidson

Department of Geology and Geological Engineering, Carrier Hall, The University of Mississippi, University, MS 38677 USA

I C Yang US Geological Survey, MS-421, Lakewood, CO 80225 USA

ABSTRACT. The radiocarbon activity of carbon collected by vacuum distillation from a single partially saturated tuff began to decline after approximately 60% of the water and carbon had been extracted. Disproportionate changes in 14C activity 613C and during distillation rule out simple isotopic fractionation as a causative explanation. Additional phenomena such as matrix diffusion and ion exclusion in micropores may play a role in altering the isotopic value of extracted carbon, but neither can fully account for the observed changes. The most plausible explanation is that distillation recovers carbon from an adsorbed phase that is depleted in 14C relative to DIC in the bulk pore water.

INTRODUCTION

Vacuum distillation was first introduced by Davidson (1995) and Davidson et al. (1995) as a method for extracting dissolved inorganic carbon (DIC) from the pores of partially saturated rock for 14C analysis. The distillation method was initially developed in order to obtain DIC from rock pores where the total water content was too low to obtain water by conventional compression or centrifugation methods (Peters et al. 1992). These studies were performed on cores from the Apache Leap Tuff in central Arizona.

The Apache Leap studies compared the 14C activity of DIC collected by distillation of preserved core samples from the Deep Slant Borehole (DSB) with the activity of CO2 in formation air samples collected from 1-m intervals at the same depth, and from DIC extracted from water collected from adjacent core samples by core compression. The measured 14C activities of samples collected by different methods near the same depths were similar (Figure 1), suggesting that the distillation method could be used to obtain representative samples for 14C analysis. 14C Minor fractionation of was anticipated during distillation of the DSB cores. Calculated '4C fractionation, based on measured b13C fractionation, was less than 4%. This figure was consistent with the differences observed in the measured 14C activities when comparing different methods of extracting carbon (Figure 1).

In the current study, DIC was collected from a preserved core from the Topopah Spring Tuff at Yucca Mountain, Nevada. Analyses of incremental carbon samples collected from the same core over time during the distillation showed a decreasing trend in the 14C activity that was not anticipated based on stable carbon isotope measurements.

METHODS

Distillation of DIC was performed on a single preserved, 3-kg (dry weight) core sample from the Topopah Spring Tuff from borehole UZW SD-12, Yucca Mountain, Nevada. The core was taken from a densely welded, devitrified, nonlithophysal zone at a depth of 211 m (692 feet). Porosity and initial saturation of the core were approximately 1.0% and 92%, respectively (Rautman and Eng- strom 1996). Preservation had been completed in the field at the time of drilling in July of 1994 by

141 142 G R Davidson, I C Yang

-'- Distillation ---D- Compression Formation Gas

20 t

40 t

60 t

80 t

100 t Figure 1 Data from Davidson et al. (1998) showing the 14C activity of 1) carbon extracted from rock 120 measured pores by distillation, 2) DIC from rock pore water collected 95 75 80 85 90 by core compression, and 3) CO2 in formation gas collected pMC from 1-m intervals

sealing the core inside a Lexan® liner and then again inside a heat-sealable aluminum laminate (Pro- tec Core®)1 Distillation followed the method outlined in Davidson et al. (1995). In summary, the preserved core sample was opened and immediately transferred into an argon-filled stainless-steel cell. The cell was cell was heated to mounted on a scale with 1 g resolution and connected to a high-vacuum line. The 180 °C and extracted gases drawn through a series of cryogenic traps to capture water (dry-ice/alco- hol traps) and carbon dioxide (liquid nitrogen traps). Total distillation time was 12 h. The mass and isotopic composition of carbon collected over time was monitored by periodically subliming the collected CO2 into a known volume with a pressure transducer followed by removal for isotopic analy- S13C 14C. sis. Incremental CO2 samples were analyzed for both and The rate of water removal was monitored by attributing mass loss over time to water loss. 14C results are reported as percent modern carbon (pMC) with an analytical precision of about 1.0% (26); b13C results are reported as per mil shifts relative to the Vienna Pee Dee Belemnite (VPDB) isotopic standard with an analytical precision of about 0.2% (26) for all samples.

RESULTS AND DISCUSSION Figure 2A shows the b13C and 14C results for the isotopic composition of each incremental carbon sample. Initial increases in b13C over time, representing the first 3 h of distillation and 80% of the 12C total carbon mass recovered, reflect a continuous enrichment of 13C in the aqueous reservoir as

i The use of brand/company/trade names is for identification purposes only and does not constitute endorsement by the U.S. Geological Survey. '4C in Rock Pore Water 143

A 80 --6 f- C-14 70 -0--- C-13 _g -10 f-- -12 50 f - -14 -16 40 -18 _nn 30 -22 20 111111111111 -24 0 1 2 3 4 5 6 7 8 910111213

B 80 -4

-6 70

-10 - -12 -14 --- -16 -18 U 30 -20

20 111111111111 -24 0 1 2 3 4 5 6 7 8 910111213 Distillation time (h) Figure 2. A. Incremental 14C activity and 8130 of carbon collected by distillation over time. B. Cumulative 14C activity and S13C of carbon distilled over time. is preferentially removed to the gas phase. Subsequent decreases in b13C probably reflect competing 13C processes where depletion in the aqueous reservoir, due to calcite precipitation, dominates over 13C continued enrichment of the gas phase through preferential loss of '2C (Davidson et al. 1995). Cumulative results (Figure 2B) show a continuous enrichment of 13C in the extracted phase until more than 90% of the total carbon was recovered. The composite sample was isotopically heavier than the initial incremental sample by 7.2%0. Although extracted carbon becomes enriched in 13C over time, the initial recovery is so depleted in 13C that the final composite sample is typically still isotopically light relative to in situ pore-water DIC (Davidson et al. 1995). 144 G R Davidson, I C Yang

due If the isotopic composition of extracted carbon over time is controlled by simple fractionation 14C in the b13C to differences in mass, the activity should follow the same general pattern observed 14C data (that is, changes in b13C and 14C should be proportional). Figure 2 shows that the activity the followed a distinctly different path. The first 2 incremental samples (Figure 2A), representing of 74.6 first 2 h of distillation and 60% of the total carbon mass recovered, had a constant activity 12 h of pMC. Each additional incremental sample decreased in activity to a low of 27.4 pMC after 14C distillation. Figure 2B shows a steady depletion of in the cumulative sample, in contrast to the 14C enrichment of 13C shown in the b13C data. If the activity were governed by the same processes 14C of affecting the stable isotopes, and it is assumed that fractionation is approximately twice that 13C (relative to 12C), the cumulative 14C should have increased to about 76 pMC. b13C 14C Three possible processes were considered to explain the decrease in and late in the distillation: l) matrix diffusion, 2) ion exclusion, and 3) adsorption.

Matrix Diffusion In theory, decreasing 14C activity can result during distillation if carbon begins to be extracted from 14C dead-end micropores in the rock (Neretnieks 1981). This is possible if was depleted in micropores in situ prior to sampling. In a dry rock, advective flow will initially fill dry dead-end micro- 14C pores with water, but no additional water will enter the pores. As present in the micropores decays, a concentration gradient may be established between the bulk pore-solution and the micro- 14C pores. Steady-state concentrations will be established as a function of the rate of diffusion of into 14C the micropores and the rate of 14C decay. The steady-state gradient will persist in a preserved core sitting on a shelf, although the overall activity will decline with storage time. During distillation of a core, the first water and carbon to be sampled is from the larger, more easily accessible pores. Later in the distillation, water and carbon are drawn from smaller pores where the carbon reservoir may be depleted in 14C as described above. At the time when the second incremental carbon sample was collected from the Topopah Spring core, about 56% of the available water had been recovered (Figure 3). Steady-state flux of 14C from the bulk solution into micropores can be modeled using a one-dimensional transport equation with a first-order reaction term: -Dd2A(x)+kA(x)_0 , (1) dx 2

where D is the diffusion coefficient (1.185 x 10_9 m2s 1 for HCO3 in water; (Lide 1997, p 5-99), k is the decay constant for 14C (3.836 x 10-12 s 1), and A(x) is the activity at distance x into a micropore. The diffusion coefficient for HCO3 is used because it is the predominant form of DIC in pore water samples obtained from related cores. If the boundary conditions for Equation (1) are set such that there is a constant activity at the mouth of the pore (A0) and no flux is allowed at the extreme end of the pore (x = L), the solution is (Bird et al. 1960, p 532-3):

cosh) [L-x] A(x)=A0 (2) cosh[ [LI] D 14C in Rock Pore Water 145

200 - 180

160 a)

, L 140 I 0) 8 120 0 a) 100

ctj 80 t U ici -o- WerYield 60 L U ci. -0- Temperature 40 --Carbon Yield a) , I F- 20

0 0 1 2 3 4 5 6 7 8 910111213

Distillation Time (h)

Figure 3. Temperature and cumulative water and carbon yield during distillation

If it is assumed that the final carbon drawn from the distilled rock came from the extreme ends of 14C micropores, the activity of the last incremental carbon sample can be represented by A(x) = 27.4 pMC where x = L. If it is further assumed that the first 60% of the carbon recovered represents the composition of the bulk pore-solution, then Ao = 74.6 pMC. Solving for L then gives an unrealistic mean pore length of 29 m. This model is imperfect because carbon distilled from micropores near the surface of the core and from the bulk solution closer to the center of the core may be recovered simultaneously. Such mixing, however, would elevate the activity of the later samples, requiring even longer pore lengths to account for the lower activity of the carbon coming from the micropores alone.

The model is also imperfect because it assumes that the pores were fully saturated. The degree of saturation in the sampled core was about 92% (Rautman and Engstrom 1996). This means that water in some of the micropores in situ may have been separated from the bulk solution by air pockets. Significant depletion of 14C in these micropores relative to hydraulically connected micropores is unlikely, however, because isotopic exchange readily occurs between gaseous CO2 and aqueous DIC (hook et al. 1974), and diffusion of CO2 in air is more rapid than diffusion of HC03 in water (Lide 1997, p 5-99 and 6-218).

Ion Exclusion

The isotopic composition of DIC in the smallest pores could be altered by interactions between charged DIC species and mineral surfaces. In this model, DIC anions (HC03 and CO)3are inhib- ited from passage into the smallest pores, resulting in an enrichment of the uncharged species (H2CO3) in the micropores. This will result in a depletion of 13C in the DIC within the micropores relative to the bulk solution, assuming equilibrium isotopic fractionation between the three DIC species (Defines et al. 1974). Samples collected late in the distillation may reflect collection of this isotopically depleted reservoir. 146 G R Davidson, I C Yang

Preferential passage of H2CO3 has been observed in hyperfiltration experiments where DIC solu- tions were forced through smectite clay (Fritz et al. 1987). The results of hyperfiltration are not directly analogous to diffusion into micropores, however, because the driving forces are not the 13C same. Fritz et al. (1987) attributed initial transient increases in DIC and enrichment of in the fil- trate to advectively driven processes: conversion of HCO3 to H2CO3 at the upstream clay-solution interface, allowing easier passage of DIC through the clay, and differential back-diffusion of the isotopes from the concentrated clay-solution layer to the bulk feed solution. b13C Ion exclusion effects may contribute to the decline in the incremental values late in the distil- 14C lation, but they do not adequately account for the decrease in activity. Fractionation induced 13C 14C, solely by ion exclusion at the micropore level should be proportional for and resulting in a much smaller decline in 14C than observed. Adsorption A more plausible possibility is that carbon sampled later in the distillation begins to draw from a secondary source of carbon, such as an adsorbed phase. If exchange between DIC and adsorbed carbon is slow, 14C in the adsorbed phase will not be readily replenished as it decays. The activity of the adsorbed phase will continue to decline until the exchange rate with the bulk solution is equal to the decay rate. In this model, carbon recovered in the early stages of distillation represents DIC degassed from the bulk solution. Later in the distillation, carbon becomes increasingly derived from the 14C-depleted adsorbed phase. The existence of a sizeable reservoir of adsorbed inorganic carbon on mineral surfaces has been suggested to explain high uptake of CO2 on experimental soils (Striegl 1988; Striegl and Armstrong 1990), and high recovery of carbon from distilled tuff cores from Superior, Arizona (Davidson et al. 1995). In these studies, the mass of carbon removed by soils or recovered from cores ranged from 3 to 17 times higher than the mass of DIC in the pore water of the samples. The existence of an adsorbed phase in the present study can only be approximately evaluated because there was no way to directly measure the original mass of DIC in the pore water of the distilled core. The total water content was too low to obtain pore water samples by typical compression or centrifugation methods. Simply dividing the total mass of carbon recovered (9.64 mg) by the total mass of water distilled (0.125 L) gives a C concentration of 77 mgIL (391 mg/L HCO3 ). This represents only a portion of the available carbon in the rock pores because some of the original mass is left behind as precipitated carbonate during the distillation. An earlier study investigating the recovery efficiency of the distillation method found that 40% to 85% of the available carbon in 6 tuff samples was left behind following distillation (Davidson et al. 1995). No DIC concentration data are available from the Topopah Spring Tuff, but concentrations of pore water obtained by compression from adjacent units (Calico Hills and Prow Pass Tuff) range from 114 to 323 mg/L HCO3 , with pH values of 8.7 and 8.3, respectively (Yang et a1.1998). The partial mass of carbon collected from the distilled Topopah Spring sample exceeds the highest total DIC concentration measured in the Calico Hills and Prow Pass samples by more than 20%. Comparison with adjacent units is speculative, but the high carbon yield is consistent with the hypothesis that carbon has been drawn from an adsorbed phase. The high yield cannot be accounted for by atmospheric leakage into the distillation cell because a leak would result in an increase in the 14C activity with time rather than the observed decline. The continued decline in S13C during the final stages of distillation is also consistent with adsorption-induced fractionation observed by others. Yang et al. (1985) collected CO2 from a subsurface formation by pumping borehole gas through a molecular sieve (MS) packed with a silicate substrate 14C in Rock Pore Water 147

and through a solution of potassium hydroxide (KOH). The S13C values of the MS samples were 5%o to 6%0 lighter than the KOH samples, suggesting preferential adsorption of 12C02 over 13C02 by 12C02 the MS. Preferential adsorption of is also routinely observed by colleagues at the University of Mississippi during gas chromatography experiments (Jon Parcher, unpublished data). The sub- strates in both the MS and chromatography experiments are not the same as in the tuff sample, however, so further investigation is warranted to measure the fractionation associated with adsorption on specific minerals found in the tuff.

The rate of carbon and water collection over time was analyzed to determine whether recovery of an adsorbed carbon phase late in the distillation would result in an increasing ratio of recovered carbon/ water (Figure 4). The opposite trend was actually observed; the ratio of recovered carbon/water decreased during the later stages of distillation. There are at least two possible explanations. First, DIC removal may be more efficient than water removal under conditions of low pressure and elevated temperature. Under these conditions, a disproportionate mass of DIC may degas early in the distillation, resulting in a diminished supply of carbon as water continues to vaporize slowly. Sec- ond, lower carbon/water yields may be expected if the later stages of distillation draw water from ion-deficient micropores (Peters et al. 1992).

110 -- Incremental 100 -o- Cumulative 90

0 2 4 6 8 10 12 Time (h)

Figure 4. The ratio of carbon/water recovered during distillation. Error bars are left off the cumulative data to avoid cluttering the figure, but are of the same magnitude as the corresponding incremental data.

CONCLUSION

14C The activity of carbon distilled from a tuff core began to decline after approximately 60% of the water and carbon was extracted. This phenomenon cannot be attributed to simple isotopic fractionation because changes in the 14C activity and S13C of recovered carbon over time followed distinctly different pathways. Matrix diffusion and ion exclusion may be active in situ processes, but neither can account for the observed decline in 14C activity. Matrix diffusion into dead-end micropores requires unrealistic pore lengths to account for the final '4C activities. Ion exclusion may contribute to the late-stage decline in b13C, but does not account for the disproportionate decline in 14C. The 148 G R Davidson, I C Yang

that is most plausible explanation is that distillation recovers carbon from an adsorbed phase 14C situ depleted in 14C relative to DIC in the bulk pore water. Depletion of may result from slow in exchange between the adsorbed phase and the bulk pore water. Viewed in isolation, these results would suggest that the distillation method will not provide carbon with representative 14C activity because the cumulative activity changed by 12 pMC over the course tuff of the distillation. Comparisons, however, between different methods of extracting carbon from have shown that the activities of distilled samples are reasonably close to those of samples obtained from adjacent depths by other methods (Figure 1). It is probable that distillation produces carbon The with a 14C composition representative of pore-water DIC within a finite range of accuracy. range is probably less than ±10%, but additional work is needed to refine this estimate.

ACKNOWLEDGMENTS Stable and radioisotope analyses were performed at the NSF-Arizona Accelerator Mass Spectrom- eter (AMS) facility at The University of Arizona.

REFERENCES of pore water ex- Bird RB, Stewart WE, Lightfoot EN. 1960. Transport liminary study of the chemistry compression. In: phenomena. New York, John Wiley & Sons. 780 p. tracted from tuff by one-dimensional interac- Davidson GR. 1995. Geochemical and isotopic investi- Kharaka YK, Maest AS, editors. Water-rock Symposium gation of the rate and pathway of fluid flow in par- tion: proceedings of the 7th International City, Utah, 13-18 tially-welded fractured unsaturated tuff [dissertation]. on Water-Rock Interaction; Park 741-5. Tucson: University of Arizona. 208 p. July 1992. Rotterdam: A. A. Balkema. p. Geology of the USW Davidson GR, Bassett RL, Hardin EL, Thompson DL. Rautman CA, Engstrom DA. 1996. Nevada. Sandia 1998. Geochemical evidence of preferential flow of SD-12 drill hole, Yucca Mountain, Laboratories, water through fractures in unsaturated tuff, Apache Report SAND96-1368, Sandia National Leap, Arizona. Applied Geochemistry 13:185-95. Albuquerque, New Mexico. 132 p. of 14-carbon Davidson GR, Hardin EL, Bassett RL. 1995. Extraction Striegl RG. 1988. Exchange and transport Madi- of 14C from pore water in unsaturated rock using vac- dioxide in the unsaturated zone [dissertation]. uum distillation. Radiocarbon 37(3):861-74. son: University of Wisconsin. 149 p. Carbon dioxide reten- Deines WG, Langmuir D, Harmon RS. 1974. Stable car- Striegl RG, Armstrong DE. 1990. Quaternary bon isotope ratios and the existence of a gas phase in tion and carbon exchange on unsaturated Acta 54: the evolution of carbonate groundwaters. Geochimica sediments. Geochimica et Cosmochimica et Cosmochimica Acta 38:1147-64. 2277-83. JN.1998. Fritz S, Hinz DW, Grossman EL. 1987. Hyperfiltration Yang IC, Yu P, Rattray GW, Ferarese JS, Ryan the induced fractionation of carbon isotopes. Geochimica Hydrochemical investigations in characterizing Nevada. US et Cosmochimica Acta 51:1121-34. unsaturated-zone at Yucca Mountain, Investigation Re- Lide DR, editor. 1997. CRC Handbook of Chemistry and Geological Survey Water-Resources Physics. 77th ed. Boca Raton (FL), CRC Press. port 98-4132.57 p. 1985. Mook WG, Bommerson JC, Staverman WH. 1974. Car- Yang IC, Herbert HH, Weeks EP, Thorstenson DC. iso- bon isotope fractionation between dissolved bicarbon- Analysis of gaseous-phase stable and radioactive Ne- ate and gaseous carbon dioxide. Earth and Planetary topes in the unsaturated zone, Yucca Mountain, Well Science Letters 22:169-76. vada. In: Proceedings of the National Water and Mon- Neretnieks I. 1981. Age dating of groundwater in fis- Association conference on Characterization 19-21 No- sured rock: influence of water volume in micropores. itoring of the Vadose (Unsaturated) Zone, Water Well Water Resources Research 17:421-2. vember 1985, Denver. Denver: National Peters CA, Yang IC, Higgins JD, Burger PA. 1992. A pre- Association. p. 488-506. RADIOCARBON, Vol 41, Nr 2, 1999, p 149-156 ©1999 by the Arizona Board of Regents on behalf of the University of Arizona

RADIOCARBON AGE ANOMALIES IN LAND SNAIL SHELLS FROM TEXAS: ONTOGENETIC, INDIVIDUAL, AND GEOGRAPHIC PATTERNS OF VARIATION

Glenn A Goodfriend) G Lain Ellis2 L J Toolin3

ABSTRACT. Accelerator mass spectrometric (AMS) radiocarbon analyses of live-collected, prebomb samples of shell carbonates of the land snails Rabdotus dealbatus and R. alternatus from Texas were carried out to quantify the characteristic age anomalies of land snails from limestone areas. Age anomalies are similar for the two species; they average +700 yr and vary by ±180 yr (la) among samples. Serial analysis of 1 shell reveals a significant ontogenetic trend in 14C age anomalies, with older apparent ages (up to 1200 yr) in the apical part of the shell and younger and uniform ages in the last whorl. No trend in age anomalies was found across a broad range of rainfall conditions (from 300 to 1000 mm mean annual rainfall).

INTRODUCTION

A number of earlier studies have established that land snail shell carbonate typically has a radiocarbon age anomaly: the measured 14C age is older than the actual age, as a result of ingestion of old carbonate (e.g., limestone) and its incorporation into the shell (Evin et al. 1980; Burleigh and Kemey 1982; Goodfriend and Stipp 1983; Goodfriend and Hood 1983; Goodfriend 1987). The ingested carbonate dissolves in the stomach acid, producing C02, which dissolves in the body fluids to become part of the bicarbonate pool. This bicarbonate pool is the source material from which calcium carbonate is precipitated during the process of shell growth (Goodfriend and Hood 1983).

In recent years, accelerator mass spectrometric (AMS) analysis of samples for 14C has become common. Because the amount of sample required for analysis of carbonates (12% C by weight) by AMS is only about 10-20 mg, it is now possible to analyze even rather small individual shells or parts of larger shells. Conventional analysis (13-decay counting) requires about 1000 times this amount, so that much of the previous work on fossil land snails involved analysis of bulk samples (i.e., comprised of many individuals) or whole shells of very large land snails. Previous studies on modem land snail shells, aimed at quantifying 14C age anomalies, were also based primarily upon analysis of bulk samples or very large individual shells. Such analyses have been used to compare variation in age anomalies among species or between different regions (e.g., Goodfriend and Stipp 1983). But how much variation in age anomalies occurs among individuals of the same population or within an individual shell as it grows? For AMS analyses, it is important to establish to what degree such variation may contribute to the overall variability of age anomalies in order to quantify uncertainties of 14C corrected ages and also to develop a sampling strategy to minimize such variability.

Shells of the land snail Rabdotus are abundant in archeological sites as well as in alluvial and cave deposits throughout much of the southern Great Plains. They have been widely used for 14C dating (Stafford 1993; Ellis and Goodfriend 1994; Ellis et al. 1996; Abbott et al. 1995; Abbott et al. 1996; Toomey and Stafford 1994; Quigg et al. 1996; Johnson forthcoming). Thus, it is important to establish the age anomalies in these snails on a quantitative basis so that accurate results may be obtained from 14C dates on fossil samples.

In the present study, we looked at 14C age anomalies in 2 species of sympatric land snails of the genus Rabdotus (dealbatus and alternatus) in Texas, based on analysis of modern, prebomb shells (i.e., shells collected alive before the thermonuclear bomb tests of the late 1950s significantly raised 14C atmospheric levels) obtained from museum collections. For 1 shell, we examined ontogenetic

(Department of Geology, George Washington University, Washington, DC 20052 USA 2Environmental Affairs Division, Texas Department of Transportation, 125 E 11th St, Austin, TX 78701-2483 USA 3NSF AMS Facility, University of Arizona, PO Box 210081, Tucson, AZ 85721 USA

149 150 GA Goodfriend, G L Ellis, L J Toolin variation within the shell by analyzing a series of samples taken from the upper whorls (the earliest part of shell growth) to the lip (the last part of the shell to be deposited). We examined variation the 2 among individual shells from the same collection and differences in age anomalies between species. We also evaluated possible geographic trends in age anomalies in relation to environmental gradients. Our samples of Rabdotus come from environments ranging from moist eastern Texas, western receiving about 1000 mm mean annual rainfall and supporting deciduous forest, to dry (Figure 1). Texas, receiving only 300 mm rainfall and supporting a sparse desert scrub vegetation

600 1200 .. L _ 800 1000 400\

200

0 100 200 300 MEXICO KM

Figure! Map of Texas, showing locations of samples analyzed and mean annual rainfall isohyets (in mm) (see Table 2 for site abbreviations)

MATERIALS AND METHODS Live-collected, prebomb Rabdotus shell samples were obtained from the collection of the US National Museum of Natural History. Individuals from 10 samples of snails were analyzed, including 7 samples of R. dealbatus4 and 3 samples of R. alternatus. The samples come from 9 localities, with both species obtained from one of the sites. Prebomb shells are used for these analyses in order 14C to maximize the precision of the age anomaly estimates. Since postbomb atmospheric levels

by Pilsbry 4We have included R. mooreanus under R. dealbatus. R. mooreanus was considered a subspecies of R. dealbatus anal- (1946). It was tentatively elevated to specific status by Fullington and Pratt (1974) but without supporting anatomical and that ysis. We consider that determination of the correct status of this form awaits detailed anatomical and genetic study available information on shell morphology does not warrant specific recognition of this form. 14C Age Anomalies in Land Snail Shells 151

vary significantly from year to year and also geographically, it is difficult accurately to estimate reference atmospheric 14C levels for postbomb snails that live for longer than 1 year, as Rabdotus does. Individual shells were usually sampled by cutting a strip of shell from roughly the last third of the shell (perpendicular to the lip and the growth lines), using a Dremel tool with a 1-inch circular saw blade. The strips were cleaned by sonication in distilled water, then dried. Such a sample is representative of the later phase of growth of the shell. In order to obtain representative values for each a of series of localities across the rainfall gradient, samples were pooled from 3 to 6 individuals. For 1 sample, each of 5 shells was analyzed individually to examine among-individual variation. One shell was also selected for analysis of 5 serial samples. These samples were removed as powders, using a motorized Dremel tool fitted with a dental bit.

Corrections for isotopic fractionation were made based on measurement of the b13C values. The 14C activity of the atmosphere contemporary with the snails (i.e., at the time of collection of each sample) was determined from analyses of tree rings by Stuiver and Becker (1993). To calculate the age anomaly, these apparent ages of atmospheric carbon were subtracted from the shell 14C ages.

Mean annual rainfall amounts for 1931-1960 were interpolated for the sites from which snail samples were analyzed based on data of the US Environmental Data Service (1969). RESULTS

Ontogenetic Variation

Five samples were analyzed from a specimen of Rabdotus alternatus from Eagle Pass, from the lip up to 66 mm toward the apex of the shell (as measured along the periphery of the shell). The 3 samples from the last whorl of the shell (0 to 13 mm from the lip) all show apparent 14C ages averaging 770 yr, with a standard deviation (6) of 30 yr (Figure 2). The 6 is actually less than the average analytical error (60 yr) and indicates no measurable difference between the apparent ages of these 3 samples. On the other hand, the sample at 40.5 mm from the lip shows a significantly higher 14C age of 955 ± 50 yr and the most apical sample at 66 mm shows a still higher apparent age of 1200 ± 50 yr. Thus there is a trend of decreasing 14C ages from the apex to the last whorl of the shell, with the last whorl showing uniform values.

The upper (more apical) part of the shell is laid down during the juvenile stage of the snail, whereas the last whorl is laid down as the snail is approaching adulthood. It may be that the faster growth rate associated with juveniles requires the snails to ingest more calcium (in the form of limestone) for

1400

1200 Figure 2 Variation in apparent '4C age of shell carbonate in a specimen of Rab- dotus alternatus from Eagle Pass. The 1000 positions of the samples are given in relation to their distance from the lip, as measured along the periphery of the 800 shell, so that the lip sample is plotted at 0 mm and the samples nearer to the apex 600 1 of the shell are to the right of the figure. 0 20 40 60 80 Error bars are the ±16 uncertainties of 14C Distance from lip (mm) the ages. 152 G A Goodfriend, G L Ellis, L J Toolin

to provide the their shells, whereas in the more mature snails, the diet of plant material may suffice taking sam- smaller amount of calcium needed for slower shell growth. For Rabdotus, it seems that Furthermore, ples for 14C analysis only from the last whorl of the shell will minimize age anomalies. a consistent sampling position should lead to more uniform results.

Variation among Individuals from the Same Population Besides the specimen on which serial analyses were carried out, 4 additional shells of R. alternates 14C col- from Eagle Pass were analyzed for apparent ages to examine variation among individuals 1) is lected from the same place at the same time. The mean apparent age of these 5 samples (Table than intrinsic variability of 855 yr, 6 =155 yr. Part of this 6 is the result of analytical error, rather ana- age anomalies. To obtain this intrinsic component of variability, the variance (s2) of the average a net 6 of 135 lytical error (6 = 70 yr) was subtracted from the overall variance (1552). This leaves 100 yr. How- yr. This indicates that variation in age anomalies among individuals is on the order of The variabil- ever, 4 of the 5 shells (A-D) show relatively uniform ages (mean of 790 yr, 6 = 85 yr). the vari- ity among these 4 shells is little more than their average analytical error (65 yr). So most of high age ability observed among the shells is the result of a single individual (E) with an unusually anomaly. 14C Table 1 Variation in age anomalies among individual shells of Rabdotus alternatus from Eagle Pass Apparent 14C age Specimen Lab nr b13C BP) Aa - -10.1 55 B AA-20602 - 95 C AA-20603 -9.8 65 D AA-20604 -5.7 ±50 E AA-20605 -9.8 85 Mean 855 a 155 Net 6 b 135

aMean value for 3 samples from the last 1l3 whorl of the shell (Figure 2) bNet a is the 6 after removal of the average analytical error (70 yr)

Although not large, the among-individual variability does contribute some uncertainty to the age anomaly correction factor. For this reason, bulk analyses of many individuals actually have an advantage over AMS analyses of individual shells as they tend to mask the among-individual variation. Characterization of age anomalies for comparison of species and evaluation of geographic trends was therefore based on analysis of pooled samples of 3 to 6 individual shells.

Differences in Age Anomalies between Species

Table 2 gives the results of analyses of 7 R. dealbatus samples and 3 R. alternatus samples; the 2 is neces- species were collected together at 1 locality (Eagle Pass). In comparing these samples, it 14C sary to compare the age anomalies (the deviation of apparent ages from contemporary atmospheric 14C ages) rather than the apparent ages themselves, since the samples were collected at different times when the atmospheric 14C levels were different (Table 2). No difference in the mean age anomaly is seen between the 2 species (mean ± standard error; +680 ± 65 for R. dealbatus; +715 ± 160 for R. alternatus). Samples of the 2 species from Eagle Pass do differ in apparent age by 235 yr, Table 2 14C age anomalies in modern, prebomb Rabdotus land snail shells from Texas. N is the number of individuals pooled for the analysis (Eagle Pass R. alternatus analyzed individually; see Table 1).

LoCalitya 14C Year of Measured atmospheric anomaly (abbreviation) Species N BP ± l) nr nrb (yr BP)° Austin (AU) dealbatus 6 50 Waco (WA) dealbatus 5 60 Lampasas (LA) dealbatus 4 ±40 Fredericksburg (FR) dealbatus 6 ±70 Columbus (CO) dealbatus 5 ±55 Eagle Pass (EP) dealbatus 5 1910 70 Dryden (DR) dealbatus 5 50 Eagle Pass (EP) alternatus 5 1910 70 Zapata (ZA) alternates 5 1932 40 Pecos R. (PE) alternates 3 1890 35

Mean +690 6 190 Net 6d 180 aAustin and Waco data from Ellis et al. (1996); Lampasas and Fredericksburg data from Quigg et al. (1996). bus National Museum of Natural History catalog number. cInte olated from Table 1 of Stuiver and Becker (1993) based on year of collection. dNet 6 is the 6 after removal of the average analytical error (54 Yr).

w 154 GA Goodfriend, G L Ellis, L J Toolin

data but this difference is similar to the average 6 (190 yr) seen in the overall data set. The available thus show no difference in age anomalies between these 2 species of Rabdotus.

Possible Geographic Trends in Age Anomalies in Relation to Rainfall When the 14C age anomalies of the Rabdotus samples are plotted in relation to the mean annual rainfall of the sites from which they were collected, no trend of ages is apparent (r=0.15) (Figure 3). Rainfall is the predominant environmental variable through the region and is related to vegetation type (including forest, prairie, woodland, and desert scrub within the region) and density, as well as to soils. Rainfall also directly affects the activity and growth rates of snails, which in turn may affect calcium requirements. Despite this wide range of environmental conditions represented by the samples, the age anomalies are not obviously influenced by them.

1200

R. dealbatus 1000 o R. alternatus

800 c0 ('3 600 Figure 3 14C age anomalies of Rabdotus 4 0) 4 dealbatus and R. alternatus shells in re- Q 400 lation to mean annual rainfall (ages from Table 2). See Figure 1 for locations of 200 400 600 800 1000 samples. Mean annual rainfall (mm)

The substrate of central and southern Texas consists mostly of limestone or sediments derived from limestone regions. Since the samples come from a museum collection and generally have only the name of the town indicated, rather than more precise locality information, it is not possible to tell in which cases samples may have been collected on alluvial sediments where snails would not have access to limestone bedrock. The Columbus site on the coastal plain is on Quaternary alluvium, the upland source of which lies on limestone. The Zapata site lies on sandstone terrain but shells could be from an alluvial context. The area around Lampasas has both sandstone and limestone bedrock. At localities where limestone bedrock is not accessible to snails, carbonate-containing sediments (such as alluvium or eolian sediments) may be utilized. If the carbonate in these sediments is of relatively recent origin (derived from soil carbonates or freshwater precipitates, rather than from lime- 14C stone clasts), then consumption of these by snails would produce a reduced age anomaly, depending on the apparent age of the carbonates.

Correcting Fossil Rabdotus 14C Ages for Age Anomalies The analyses presented above indicate that modern prebomb Rabdotus dealbatus and R. alternatus 14C shells have an average age anomaly of +700 yr. In order to correct ages of fossil Rabdotus for age anomalies, this amount should be subtracted from the apparent ages, before calibration. Because the age anomalies do not appear to depend on climatic conditions such as rainfall (based on modern geographic variation), this correction should be applicable to fossil samples, even if the past cli- mates they lived under were different from the modern climate. The age anomalies are variable, however, and this must be taken into account in calculating the overall error of the corrected 14C ages. Among-sample variation (based on bulk samples of about 5 14C Age Anomalies in Land Snail Shells 155

shells) is estimated at 180 yr (6). This would be the appropriate number to combine with the analytical error of a fossil sample (the variances sum) to obtain the error of the corrected age, if the fossil sample is also a bulk sample of about this number of individuals. However, if the fossil sample consists of only 1 shell, then there will be a slightly larger variability of the age anomaly, because among-individual variation exceeds analytical error (net a of 135 yr). To calculate what the overall error should be for a single shell, we must first calculate the expected among-sample component of the variability by subtracting the average standard error of each sample (assumed to be represented by the multiple analyses of R. alternatus individuals from Eagle Pass, or 135/J5 = 60 yr) from the net 6 of 180 yr (I(1802 - 602) =170 yr). To this value of 170 yr for the among-sample component of variation, we must add the among individual variation (6 =135 yr) to obtain the overall error of the age anomaly correction for an individual shell analysis: J(1702 + 1352) = 210 yr. This error is slightly larger (by 30 yr) than the error for bulk samples (for n=5). To obtain the overall error of the 14C corrected age (6total) for a fossil sample, the error of the age anomaly must be combined with the 14C analytical error for the fossil sample as reported by the lab (61ab) Thus the overall errors may be calculated according to the following formulae:

for individual shell: 6total = 2102 + 61ab (1)

for bulk analyses (e.g., N=5): 6total = J1802 + 62 (2)

Because of the variability of the age anomaly, corrected 14C dates on Rabdotus shells can never be more precise than approximately ±200 yr. The analytical error for Holocene fossil samples is generally much smaller than this (typically 20 to 80 yr) and thus does not contribute very much to the overall error. However, in older snail samples in which analytical errors are on the order of several hundred years, the variability of the age anomaly contributes relatively little to the overall error of the corrected ages.

It should be noted that these correction factors apply to samples taken from the last whorl of Rab- dotus shells. Pieces of shell from the apical whorls of the shells are expected to have higher age anomalies, based on the observed pattern of ontogenetic variation. For analyses of whole shells, the age anomaly correction should be similar to or slightly higher than the correction for samples from the last whorl, since the last whorl comprises the bulk of the shell.

The variability of age anomalies for Rabdotus reported here (6 = 180 yr) is smaller than that reported earlier for land snails from the Negev Desert (6 = 230 yr for Trochoidea and 500 yr for Sphincterochila) and much smaller than that reported for Jamaican Pleurodonte (6 = 1200 yr) (Goodfriend 1987). Thus Rabdotus is a relatively good land snail taxon for dating. Its age anomalies are also smaller than those for most other snails (usually in the range of 1000 to 2500 yr; see references in first paragraph of Introduction). This relatively small average age anomaly of +700 yr implies that only a small proportion of the shell carbonate carbon of Rabdotus derives from ancient carbonate (limestone) sources (8%, calculated according to Goodfriend 1987, Equation 2). Conse- quently, the standard correction for isotopic fractionation gives a good approximation to the actual fractionation correction. It is not necessary to use the more accurate but more complicated correction (Goodfriend 1987, Equation 8), which also takes into account the contribution of limestone to the b13C value.

Materials that have been widely used for 14C dating of Quaternary deposits and archeological sites in the southern Great Plains include charcoal, soil organics, bone, and occasionally wood. Although charcoal and wood remain the best materials for 14C analysis, they are often not preserved at sites in 156 G A Goodfriend, G L Ellis, L J Toolin the region. Corrected 14C dates on Rabdotus shells will have poorer precision, with errors on the 14C order of ±200 yr. However, they are still generally preferable to soil organic analyses for dating of strata, as soil organics are easily contaminated by subsequent root growth and generally accumu- late over long periods of time (perhaps hundreds of years). Bone can be dated with better precision than land snail shells, as bone organics do not have age anomalies. However, the accuracy of bone dates depends on the degree of preservation of the bone organics (Stafford et a1.1990). If the avail- 14C able bone is not well preserved, then Rabdotus shells would be a preferable material for dating.

ACKNOWLEDGMENTS Shell samples for this study were provided by M G Harasewych, R Hershler, and P Greenhall of the National Museum of Natural History. M Quigg and C Lintz of TRC-Mariah Associates made available to us their results of modern Rabdotus 14C analyses. This study was supported by NSF grant nr 5BR9510869 to Goodfriend and Ellis.

REFERENCES sources Abbott JT, Ellis GL, Goodfriend GA. 1995. Chronomet- sis of land snail shells: implications for carbon ric and integrity analysis using land snails. In: Abbott and radiocarbon dating. Radiocarbon 25: 810-30. the prob- JT, Trierweiler WN, editors. NRHP significance test- Goodfriend GA, Stipp JJ.1983. Limestone and ing of 57 prehistoric archeological sites on Fort Hood, lem of radiocarbon dating of land-snail shell carbon- Texas. Volume 2. Archeological Resource Manage- ate. Geology 11:575-7. ment Series, US Army, Fort Hood. Research report nr Johnson L. Life and death as seen at the Bessie Kruze Site 34:801-14. (41 WMJ3) on the Blackland Prairie of Williamson and Abbott JT, Goodfriend GA, Ellis GL. 1996. Landsnail in- County, Texas. Texas Dept of Transportation vestigations. In: Trierweiler WN, editor. Archeologi- Texas Historical Commission, Austin, Texas. Forth- cal testing at Fort Hood: 1994-1995. Volume 2. Ar- coming. cheological Resource Management Series, US Army, Pilsbry HA. 1946. Land mollusca of North America Fort Hood. Research report nr 35:619-36. (north of Mexico). Volume 2, Part 1. Monographs of Burleigh R, Kerney MP. 1982. Some chronological im- the Academy of Natural Sciences of Philadelphia 3., plications of a fossil molluscan assemblage from a Philadelphia: George W. Carpenter Fund. vii + 520 p. Neolithic site at Brook, Kent, England. Journal of Ar- Quigg JM, et al. 1996 Early archaic use of the Concho chaeological Science 9:29-38. River terraces: cultural resource investigations at Ellis GL, Goodfriend GA. 1994. Chronometric and site- 41 TG307 and 41 TG309 Tom Green County, San An- formation studies using land snail shells: Preliminary gelo, Texas. Texas Department of Antiquities Techni- results. In: Trierweiler WN, editor. Archeological in- cal Report no. 11058. Austin: TRC Mariah Associ- vestigations on 571 prehistoric sites at Fort Hood, Bell ates. 14C and Coryell Counties, Texas. Archeological Resource Stafford TW. 1993. AMS dating in geoarchaeology: Management Series, US Army, Fort Hood. Research field and laboratory controls on absolute accuracy and report nr 31: 183-201. precision. Abstracts with Programs, Geological Soci- Ellis GL, Goodfriend GA, Abbott JT, Hare PE, von Endt ety of American 1993 Annual Meeting (Boston; Octo- DW. 1996. Assessment of integrity and geochronol- ber 1993): A187. ogy of archeological sites using amino acid racemiza- Stafford TW, Hare PE, Currie, L, Jull AJT, Donahue D. tion in land snail shells: examples from central Texas. 1990. Accuracy of North American human skeleton Geoarchaeology 11: 189-213. ages. Quaternary Research 34:111-20. Evin J, Marechal J, Pachiaudi C, Puissegur JJ. 1980. Stuiver M, Becker B. 1993. High-precision decadal cali- Conditions involved in dating terrestrial shells. Radio- bration of the radiocarbon time scale, AD 1950-6000 carbon 22(2):545-55. BC. Radiocarbon 35(1):35-65. Fullington RW, Pratt WL. 1974. The aquatic and land Toomey RS, Stafford TW. 1994 Paleoenvironmental and Mollusca of Texas. Part 3. Dallas Museum of Natural radiocarbon study of the deposits from Hall's Cave, An- History Bulletin 1. Dallas: Dallas Natural Science As- Kerr County, Texas. Program and Abstracts, 65th sociation. iv + 48 p. nual Meeting of the Texas Archeological Society Goodfriend GA. 1987. Radiocarbon age anomalies in (Lubbock, Texas; November 1994): 96. shell carbonate of land snails from semi-arid areas. US Environmental Data Service. 1969. Climatology of Radiocarbon 29:159-67. the United States. Nr 60-41. Goodfriend GA, Hood DG. 1983. Carbon isotope analy- RADIOCARBON, 2, Vol 41, Nr 1999, p 157-168 01999 by the Arizona Board of Regents on behalf of the University of Arizona

CHANGE OF DIET OF THE GREENLAND VIKINGS DETERMINED FROM STABLE CARBON ISOTOPE ANALYSIS AND 14C DATING OF THEIR BONES

Jette Arneborgl Jan Heinemeier2 Niels Lynnerup3 Henrik L Nielsen2 Niels Rude Arny E Sveinbjornsdottir4

ABSTRACT. Bone samples from the Greenland Viking colony provide us with a unique opportunity to test and use 4C dating of remains of humans who depended upon food of mixed marine and terrestrial origin. We investigated the skeletons of 27 Greenland Norse people excavated from churchyard burials from the late 10th to the middle 15th century. The stable carbon isotopic composition (813C) of the bone collagen reveals that the diet of the Greenland Norse changed dramatically from predominantly terrestrial food at the time of Eric the Red around AD 1000 to predominantly marine food toward the end of the settlement period around AD 1450. We find that it is possible to 14C-date these bones of mixed marine and terrestrial origin precisely when proper correction for the marine reservoir effect (the 14C age difference between terrestrial and marine organ- isms) is taken into account. From the dietary information obtained via the S13C values of the bones we have calculated individual reservoir age corrections for the measured 14C ages of each skeleton. The reservoir age corrections were calibrated by 14C comparing the dates of 3 highly marine skeletons with the '4C dates of their terrestrial grave clothes. The calibrated ages of all 27 skeletons from different parts of the Norse settlement obtained by this method are found to be consistent with available historical and archaeological chronology. The evidence for a change in subsistence from terrestrial to marine food is an important clue to the old puzzle of the disappearance of the Greenland Norse, obtained here for the first time by measurements on the remains of the people themselves instead of by more indirect methods like kitchen-midden analysis.

INTRODUCTION

Bone Dating

14C The dating of bone is by now technically well established, relying on refined chemical extraction techniques combined with accelerator mass spectrometry (AMS) (for example, Brown et al. 1988). Since very small, even submilligram-sized, samples of bone collagen can be dated with AMS, it has become possible to select the best samples from a skeleton, minimizing problems with degradation and contamination. If the bone is reasonably well preserved, AMS 14C ages as well as stable carbon isotopic ratios can be determined reliably for skeletal remains of archaeological interest without destroying the object. If the bone collagen is of terrestrial origin, the measured (conventional) 14C age is converted into a true calendar age by using the global tree-ring calibration curve (Stuiver and Polach 1977). However, this simple procedure is not applicable when the bone collagen is derived in part from marine carbon which, due to the marine reservoir effect, appears several hundred 14C years older than the corresponding terrestrial carbon. This seriously constrains the dating of bones of people who have had access to food protein from the sea. Therefore, archaeologists have generally distrusted the precision of 14C dates of human bones. But precise 14C dating of human bones is so attractive to the archaeologist that it is highly desirable to add bone to the list of datable material. To extend the calibration of measured 14C ages to "marine" bones one needs to know both the marine food fraction and the reservoir age, that is, the age difference between the atmosphere and the particular region of the sea at the time the protein was produced.

Carbon Isotope Fractionation and Diet

Previous investigations have shown that 613C of bone collagen can be used as an indicator of food composition (b13C is the fractional deviation of the 13C/12C ratio from the VPDB standard). The

'Department of Prehistory and the Middle Ages, The National Museum of Denmark, DK-1220 Copenhagen, Denmark 2Institute of Physics and Astronomy, University of Aarhus, DK-8000 Aarhus, Denmark 3Laboratory of Biological Anthropology, The Panum Institute, University of Copenhagen, DK-2000 Copenhagen, Denmark 4Science Institute, University of Iceland, IS-107 Reykjavik, Iceland

157 158 JArneborg et al. composition determined may be either the relative components of marine/terrestrial food protein (Tauber 1981; Chisholm et al. 1982; Johansen et al. 1986) or plants of C3 and C4 photosynthesis (e.g., van der Merwe and Vogel 1978; van der Merwe 1982) and according to some authors (Chis- holm et al. 1982; Lovell et al. 1986; Chisholm et al. 1983; Heinemeier and Rud 1997), the marine b13C food fractions can even be deduced with reasonable precision (±10%) via measurements of bone collagen for individuals in a human population. The data compiled in Table 1 are for population groups from high northern latitudes where the b13C signal provides a sharp distinction between marine and terrestrial food, since C4 plants (with a different photosynthetic pathway and isotopic b13C fractionation [Lovell et al. 1986]) are not present in these areas. The distribution for a single population group can be extremely narrow (standard deviation about 0.3%0, just a few times the measuring uncertainty) (Table 1). This means that the variability in metabolic isotope fractionation b13C among individuals is negligible. Hence we conclude that differences in of human bone collagen from high latitudes must reflect real differences in the average diet consumed by the individual over roughly 10 years, which represents the collagen turnover time in human bone.

b13C Table 1 Bone collagen values for population groups from northern regions 613C %o Percent Locality Period Type N diet Tuna, Swedena Viking age Inland 7 0.26 6 Leksand I, Swedena Medieval Inland 1 Leksand II, Swedena 17th century Inland 0.33 4 Heidal, Norwaya Medieval Inland 0.3 5 Saskatchewan, Canadab Prehistoric Inland 0.3 -b British Columbia, Canada Prehistoric Coastal 0.4 91 West Greenland, Eskimosd 15th century Coastal 8 0.18 100 S13C 13C/12C N is number of individuals in the population group. is the relative deviation of the isotopic ratio from the VPDB standard. The variabilities in the 813C values are one standard deviation. Percent marine diet is calculated from 813C by linear interpolation between the 813C values -12.5%o and -21%0, taken to be the endpoint values for purely marine and purely terrestrial (pure C3) diet respectively. aThese data are for inland populations from Sweden (Liden and Nelson 1994) and Norway (Johansen et al. 1986) known to have negligible access to marine food. bThe less negative 8'3C average for this inland population is ascribed to admixtures of C4 plants in diet (Lovell et al. 1986). Chisholm et al. 1983. 813C TThis West Greenland Eskimo population were close neighbors to the Norse (Heinemeier and Rud 1997). The average is identical to our choice of endpoint value for 100% marine food.

All inland sites where C3 plants compose local vegetation, such as in Greenland and Scandinavia, show narrow b13C distributions without significant difference in absolute values (Table 1). This is consistent with the geographically and archaeologically expected well-defined (terrestrial) food pattern and leads us to adopt the endpoint S13C value of -21%o for a 100% terrestrial diet. The marine endpoint value is more difficult to establish, as no human population a priori can be expected to have a 100% marine diet. This is probably reflected in the scatter and absolute value of the prehistoric British Columbia Indian population, which is archaeologically expected to have had a dominantly marine food pattern, but probably with a non-negligible (and individually varying) terrestrial component. We have adopted the marine endpoint value -12.5%o from the results of our previous measurements (Heinemeier and Rud 1997, see Table 1) of a series of Thule Culture Eskimos who lived close to the Norsemen in location and time period (Grummesgaard-Nielsen 1997; Lynnerup et al. 1997). As they are the most "marine" individuals that we are aware of at present, we assume a 100% Change of Diet of the Greenland Vikings 159 marine diet, although only future isotopic research (for example, b15N as an indicator of trophic level) can reveal whether a significant terrestrial (e.g., reindeer) diet component was present. In any case, the known, extremely marine, economy of this Eskimo population represents a natural reference for the degree of adaptation of the Greenland Norse to a similar diet pattern. The endpoint values adopted for the present work are close to those suggested by Chisholm (1989). As it turns out, our calibrated ages are rather insensitive to the exact choice of endpoints (see below).

The Greenland Norse

The aim of the present work was to investigate the potential of using a simple linear interpolation between the endpoint (pure marine versus pure terrestrial) b13C values to 1) calculate the marine fraction of each individual, and 2) correct 14C dates of mixed marine material. Our test material is bones of Greenland Norse. The story of the Greenland Norse (Krogh 1967; Jones 1986) began around AD 985, when a group of Icelandic farmers, lead by Eric the Red, colonized Southwest Greenland. Together with other small chieftains, Eric the Red and his household founded the Eastern Settlement (see Figure 1). Another group of settlers went farther north to the area around present Nuuk, where they founded the Western Settlement. The number of Norse people in Greenland is estimated to have peaked at 4000-5000 individuals. For reasons that have been debated for years, their number declined and the settlements were finally abandoned completely, probably in the middle of the 15th century. Present knowledge of the Norse culture in Greenland is based chiefly on the available written sources and the results of archaeological excavations of the remains of farmsteads and churchyards. The excavations indicate that Norse subsistence depended on the products of animal husbandry and seal hunting. However, quantitative information on the diet of the Norse is difficult to obtain from the usual archaeological sources, since the food refuse preserved in farms and the

Settlement

Eastern Settlement Brattahlio

GarOarV' Undir Hofoa Narsarsuaq Herjolfsnes

Figure 1 Map showing the location of Norse churchyards in Greenland. Sandnes is in the Western Settlement near Nuuk, the modern capital of Greenland. Five churchyards are in the Eastern Settlement in the southernmost part of Greenland. 160 J Arneborg et al. middens may not be truly representative (e.g., Renfrew and Bahn 1991). To throw more light on this aspect of Norse history by establishing the necessary, '4C-based, chronology in the absence of other means of dating burials, it was decided to investigate the skeletal remains of the people themselves by carbon isotope analysis. The present research was initiated to help determine the corrections on '4C ages of bone collagen required because of its mixed marine and terrestrial origin of ingested foods.

METHODS AND ANALYSIS

The 14C-dating was carried out on a set of bone samples from 6 Norse churchyards in Southwest Greenland, 5 from the Eastern Settlement and 1 from the Western Settlement (Table 2). The 27 individual bones were selected from about 450 skeletons excavated between 1921 and 1981 by the Dan- ish National Museum and now kept at the Laboratory of Biological Anthropology in Copenhagen. The selection strategy was meant to cover the entire settlement period and provide some geographical variation. Care was taken to choose skeletons from graves with at least some stratigraphic information available or to choose skeletons that had been found in close stratigraphic connection with other datable finds. Unfortunately, grave goods are normally absent in Christian graves, making 14C human bone dating the only means of establishing the ages of the burials, but in several cases, instead of using wooden coffins, the Norse Greenlanders buried their dead wrapped in clothes that could be used as a dating reference. Sex and age of the buried persons were determined when possible. Samples of bones and textiles were analyzed for 14C and stable carbon isotopes (Table 2). We had to address the question of possible contamination from carbon-containing preservatives. Thus, the textiles from Herjolfsnes may have been treated at the time of excavation with Beticol, an animal collagen glue (Norlund 1924). This preservative is not expected to cause problems, since it dissolves during pretreatment (A-A-A) with IM HCl at 100 °C for 10 min. Some of the bones had visible signs (a thin, shiny film) of conservation treatment. Likely preservatives are either white joiner's glue (collagen, typically extracted from horse bone) used during early excavation in Greenland (Hansen 1924) or Bedacryl (a synthetic polymer) used more recently. Prior to chemical pretreatment, the surface of bone samples was removed by scalpel to eliminate contamination and possible preservatives. Bedacryl is insoluble in the collagen extraction process, but the collagen preservative potentially poses a problem. However, we feel confident that the mechanical surface cleaning reduces the possible contamination level to below a few percent. As an example, it would take 5% contamination and the worst case of a strongly marine influenced bone treated with horse glue to produce a shift in b13C of -0.4% and a 14C shift of 50 yr BP for a very early bone (e.g. Brattahlid) treated with modem (pre-bomb) glue. Collagen was extracted from 100-200 mg bone samples and combusted to carbon dioxide. We used a modified Longin method (Longin 1971; Brown et al. 1988) with pretreatment consisting of demin- eralization with 1M HCl at 5 °C for about 30 min, humic acid removal with 0.2M NaOH, and reflux with 0.O1M HCl at 70 °C for 16 h. Nearly all samples had collagen yields of close to 5% or above, which is a safe limit for reliable dating recommended by Hedges and van Klinken (1992) and others.

Only 1 sample was as low as 1 % and therefore rejected for dating. Most of the CO2 was converted to graphite for AMS 14C dating with the EN tandem accelerator at the University of Aarhus, and the rest was used for b13C measurement by conventional mass spectrometry at the Science Institute, University of Iceland (Table 2). Detailed archaeological and anthropological information about the samples may be found in Lynnerup (1998). To improve the accuracy, at least 2 samples were prepared from each bone specimen. All measurements were used (no outliers) and all multiple samples were in agreement within statistical error. Our results for the recent TIRI intercalibration (Gulliksen and Scott 1995) showed that the precision reflects the actual accuracy. Change of Diet of the Greenland Vikings 161

Table 2 14C dates and b'3C values for bone collagen and cloth from Norse churchyards in Greenland Lab Conventional Calendar AD number Object Sample 14C age (AAR-) ID number sex, age (yr BP) AD) VPDB) Eastern Settlement Herjolfsnesa (Eli], Ikigaat) 1271 IV/KAL1106 ±45 1290 D10606 Cloth 553 ± 45 1269 XVIII/KAL906 F, 20-25 899 ± 84 1289 D l0605 Cloth 480 ±43 1270 I/KAL 1105 F, 45-50 750 ± 56 1288 D l0581 Cloth 480 ± 60 2201 D 106l2 "Burgundy cap" 685 ±40 2200 D10594 Garment 650 ±40 Brattahlidb "Tjodhilde's Church" (E29a, Qassiarsuk) 1275 KAL1180 M,>35 1229 ±41 976 1571 KAL1054 F, 25-30 1225 ± 51 985 1273 KAL0380 Ox bone 1040± 80 1011 1267 CLA-2 M, adult 1155 ±46 1020 1268 CLA-1 M, adult 1112±51 1065-1115 1568 KAL1041 F, 35-40 997 ±51 1165 1272 KAL 1060 F, adult 980 ± 49 1169 1570 KAL1059 F,>35 1092 ± 55 1172 1569 KAL 1043 F, 35-40 985 ± 45 1175 1276 KAL1789 M,50-55 1025 ±50 1192 Garoare (E47, Igaliku) 1437 KAL0915 M,30-35 1030 ± 65 1233 1439 KAL1118 M, adult(B) 880 ± 55 1272 1438 KAL0916 F, adult 880 ± 90 1295 Undir Hofdad (E66, Igaliku kuljalleq) 1442 KAL0920 M,30-35 890 ±45 1297 1441 KAL0919 F,25-30 880 ± 55 1392 Convente (E 149, Narsarsuaq) 1265 II,1/KAL1002 F, 35-40 886 ± 48 1322 1264 1,10/KAL1001 M, adult 937 ± 53 1389 1266 I,6/KAL0999 u,15-20 852 ± 44 1399 1263 I,7/KAL1000 M,25-30 845 ±50 1404 Western Settlement Sandnesf (W5], Kilaarsarfik) 1143 KAL0929 M,35-40 1030 ± 45 1145 KAL0960 F, 40-45 940 ±45 1147 KAL0959 F, 40-45 940 ±40 1148 KAL0964 F, 25-30 970 ±40 1146 KAL0961 F, 20-25 970 ±40 1144 KAL0928 F, 20-25 865 ±40 AAR-number refers to the AMS 14C dating sample. Object ID-number is the registration number of the National Museum of Denmark, where available (followed by slash) and the registration number of the Copenhagen Anthropological Labora- tory (KAL or CLA). Archaeological sites are given Norse names on the basis of an assumed identification with locations mentioned in Icelandic history writing and sagas. Site numbers and modern Greenlandic names are given in parentheses. F = female, M = male, u = unknown sex. B = skeleton of bishop. Conventional 14C ages were converted into calendar year by using a calibration curve interpolated between the terrestrial (tree-ring) curve and the model-calculated marine curve with the fraction of marine diet as an input parameter. The 813C values are with respect to the VPDB standard; the uncertainty is ±0.05% (at 16). The percentage of marine diet is calculated by linear interpolation between the end-point values -12.5% (100% marine) and -21% (100% terrestrial). We estimate an uncertainty of 10% in the percentage value. Details of the samples and sample ID numbers may be found in Lynnerup 1998. The relevant excavations have been described in: aNorlund 1924; Hansen 1924. b Krogh 1967. e Norlund 1930; Broste et al. 1944. a Broste et al. 1944. e Vebaek 1991. f Roussell 1936; Fischer-Moller 1942. 162 J Arneborg et al.

The human bones display extreme variation in b13C values (-14.5 to -19% VPDB) (Table 2), nearly spanning the full range between the marine and the terrestrial values of Table 1. When the corresponding fraction of marine food consumed by each individual is calculated from the measured S13C value by linear interpolation between the adopted endpoint values (see above), the values fall in the range of 20 to 80% marine diet (Table 2). Information on the marine reservoir effect as needed for the 14C age calibration is available from recent model calculations of the damped response of the sea to the varying 14C level in the atmosphere. We have used the global marine model by Stuiver, Pear- son, and Braziunas (Stuiver et al. 1986), later revised (Stuiver and Braziunas 1993) and implemented (Stuiver and Reimer 1993) in the 14C age calibration program CALIB 3.03. This program calculates the calibration curve valid for a given fraction of marine food by linear interpolation between the terrestrial tree-ring curve and the smoother marine curve calculated from a global model of the mixed- layer of the world ocean. The 100% marine curve used for interpolation is shifted relative to the standard curve by a constant, AR (14C years), which accommodates the deviation in reservoir age of the local sea from that of the model world ocean, which varies in time around a value of 400 yr. The AR value was determined in a calibration procedure by using 3 bone/cloth sample pairs, assumed to have pair-wise identical true age.

The first 6 samples in Table 2 represent these 3 skeletons (AAR-1271, -1269 and -1270) and their respective grave clothes (AAR-1290, -1289 and -1288) as excavated (in 1921) from a group of burials found in a corner of the churchyard at Herjolfsnes (Norlund 1924). The bodies were positioned on top of each other, thus believed to be buried at nearly the same time. While the 14C dates of the 3 cloth samples coincide within the measuring uncertainty, the dates of the bones at first sight showed a surprisingly large scatter (Figure 2). For example, skeleton AAR-1269, a young woman, was 420 14C years older than her grave clothes. This apparent age difference is explained by the marine reservoir effect, since the S13C value of her bone showed that her average food during the approximately 10-yr turnover time of carbon had been highly marine (78%), while the woolen clothes are expected to be purely terrestrial. In fact, this assumption is confirmed by the narrow distribution of the measured b13C values around an average of -22.2 ±0.2% VPDB for the 6 textiles, as hair and wool is expected to be approximately 1% more negative in b13C than bone collagen (Vogel et al. 1978; van den Merwe 1989). Thus, we see no sign of seaweed feeding of sheep (see below) in this case and, with 1 exception (see below), the cloth dates are also consistent with the archaeological dating (Arneborg 1996).

These 3 textile samples are therefore ideally suited to fix the reservoir correction parameter AR, especially since 1 of the associated bone samples (AAR-1269) is close to the maximum marine component found in the sample set. Since bone and cloth must have practically the same true age (the turnover time of carbon in the bone is assumed to match approximately the lifetime of the cloth), it is necessary to adjust the AR value of the marine calibration curve for each of the 3 bone dates until the calibrated ages coincide with that of the textile. The analysis is illustrated in Figure 2. Identical AR values of +50 yr were obtained for all 3 bone/cloth pairs, in spite of the fact that the marine fractions of the bones differ greatly. This is a crucial point that strongly supports the validity of the reservoir correction in the calibration procedure. The displacement is AR = 50 yr, corresponding to a reservoir age varying between 400 and 500 14C yr for the time range shown. Another important check on the calibration procedure is the continuation of the shifted curve to the period AD 1800- 1910, which gives values varying between 400 and 465 years. This is in good agreement with existing measurements of reservoir age for southern Greenland, all of which have been carried out on mollusks from the 19th and 20th century and yielded values in the range 400-550 yr depending on location. The fitted parameter AR is coupled to the choice of b13C endpoint values through the fix- Change of Diet of the Greenland Vikings 163

1500

1400

1300

1200 Figure 2 14C calibration curves. Terrestrial curve is based on 1100 tree-ring measurements (Stuiver and Braziunas 1993). m 1000 Marine curve (100%) is a model calculation for the mixed as layer of the ocean (Stuiver and Braziunas 1993). Curves 900 labelled 78% and 56% marine are interpolations. The data v 14C 800 points are measured ages of 3 bone/cloth pairs from the Norse churchyard at Herjolfsnes, Greenland. The cloth data 700 are plotted on the terrestrial curve. The b13C of the bone samples showed marine fractions of 55%, 56%, and 78%. By 600 assuming each bone/cloth pair to be of identical calendar age, 500 the mixed-layer curve was adjusted by an upwards parallel shift to make the interpolated curves fit the bone data points. 400 The displacement is OR = 50 yr, corresponding to a reservoir 300 age varying between 400 and 50014C yr for the time range 1000 1100 1200 1300 1400 1500 shown. Year AD points provided by the textile dates, which make the resulting calibrated 14C ages rather insensitive to the chosen endpoints.

We conclude from the above that it is reasonable to use marine calibration curves thus obtained to derive calendar ages for the whole time period studied in this work (Table 2). All calendar ages are consistent with archaeological information. In the following we discuss the ages for 4 selected burial groups in relation to independent information.

DISCUSSION OF 14C DATES

According to archaeological dates, the oldest Norse bones were found at Erik the Red's farm Brat- tahlid in the churchyard of the so-called Tjodhilde's Church (cf. Krogh 1967; also Table 2 and Fig- ure 3). The date of an ox-bone fragment (purely terrestrial) found in a mass grave proved that the Norse people and their cattle were present at Brattahlio at precisely the time (AD 985) mentioned in the Icelandic sagas for Norse arrival (landnam). Two of the human bones, a male and a female (AAR-1275 and AAR-1571), yielded similar old ages, but only after the marine correction was applied. Without the marine correction the skeletons would have antedated the arrival of Eric the Red (AD 985) by 100-200 yr, contradicting historical evidence. The human bones from Brattahlid were 20-50% marine. This scatter could be explained either by social differences in subsistence or by new immigration. Thus, low values are expected for people newly arrived from Iceland or Nor- way with a more terrestrial b13C imprinted in their bones. The youngest bone dates from Brattahlid (AD 1100-1200) have large uncertainties associated with them due to a plateau in the calibration curve (see Figure 2).

According to an account on the Nordic settlements in Greenland from the second half of the 14th century (Ivar Baroarsson's Greenland Description in Jonsson 1930), the Western Settlement was abandoned in the middle of the 14th century. From that area, we have dated 6 skeletons from the churchyard at the farm Sandnes. The skeletons were found (in 1930) in a stratigraphic sequence on top of each other, presumably buried almost simultaneously (Roussell 1936). The graves were situ- 164 J Arneborg et al. ated west of the church in an area that became part of the churchyard at a late date in the history of the site. The calibrated 14C dates of the skeletons, having high, but different, marine food fractions, do not contradict the written record, but some of the uncertainties are quite large. The dates of the 5 lowest-lying skeletons (first 5 in Table 2, Sandnes) cluster around AD 1300.

The Norse ruins at Igaliku have been identified with Garoar, the residence of Greenland's bishop (Norlund 1930). The skeleton (AAR-1439) of a bishop (buried with his crosier) was found in the north chapel of the church ruins. Directly below the bishop lay the skeletons (AAR-1437 and AAR-1438) of a male and a female side by side. The bones of the bishop were only 25% marine, perhaps indicating that he had arrived a short time before from Norway where all the bishops of Greenland originated (Arneborg 1990). The low marine content of the bishop's food could also be explained by his high social status, presumably allowing access to beef and game. The bones of the male and female skeletons had considerably higher marine food fractions, 50 and 40%, respectively. We note that the marine corrections, although widely different, lead to coinciding calibrated ages for the 3 skeletons from Gardar. An independent means of dating the skeletons relies on analysis of the arranged arm-position of the buried person, a method applicable to dating graves in Scandinavia (e.g., Redin 1976, Kieffer-Olsen 1993). Though the relative development of the positions of the arms (A: arms along the side; B: hands over pelvis; C: lower arms parallel over stomach; and D: hands over chest) have been more or less accepted, the absolute dates are still debated. In a work that involved graves from 8 churchyards in medieval Denmark (including Scania in South Sweden and Slesvig in Northern Germany), Kief- fer-Olsen (1993) concluded that arm position A dominated until ca. AD 1250, arm position B dominated ca. 1250-1350 and arm position C dominated from ca. 1350 until the mid-fifteenth century, after which arm position D became the most common. Evidently Norse Greenland graves follow the described relative development of arm positions (for more details, see Lynnerup 1998, p 55 ff); still, it would not be reasonable uncritically to apply South Scandinavian chronology to the Norse Green- land situation. Nevertheless, we note that the 3 dead persons in the north chapel of the Garoar church (E47) (Norlund 1930) were buried with their arms in position B (ca. AD 1250-1350), which is consistent with the calibrated ages for the 3 skeletons (AAR-1437, AAR-1439 and AAR-1438). Simi- larly at the convent in Narsarsuaq (E149) (Vebaek 1991), the arm position B (ca. AD 1250-1350) is represented in 2 cases (AAR-1265 and AAR-1264), and arm position C (ca. AD 135-1450) likewise in the 2 other cases (AAR-1266 and AAR-1263).

We conclude from this discussion that the pattern of 14C dates for bones of mixed marine/terrestrial origin is consistent with other age evidence. We observe repeatedly that groups of skeletons, which are likely to have been buried at the same time, come out with identical calibrated 14C ages (within experimental uncertainty) even in cases where the marine reservoir corrections differ markedly. We also observe accordance between the archaeological dates and the calibrated 14C ages. It thus appears that mixed marine bone samples can be reliably dated with reservoir corrections based on measured collagen b13C values.

Greenland may constitute a particularly simple system in this respect due to the absence of calcareous freshwater sources, which are known to produce high local reservoir effects of more than 1000 yr, for example at Dutch sites (Lanting and van der Plicht 1996) and Danish fjords (Heier-Nielsen et al. 1995). Norse Greenland sheep and cattle may have had a small marine component from seaweed and fish refuse, as shown for recent times in Norway (Hoegh 1982) and Iceland (Kristjansson 1982). Because the corresponding 13C isotope value will be passed on to the consumer, the reservoir correction procedure will still be valid. Questions remain about the effect of the high variability in, for Change of Diet of the Greenland Vikings 165

example, Greenland seal bone b13C values, often "less marine" than the Eskimos in Table 1, observed from measurements in our own laboratory and reported by other authors (e.g. Tauber 1984). If these values in bone collagen reflect anomalous b13C values in seal meat consumed by humans under study, it could lead to errors in evaluated marine food fractions. A similar ambiguity in the calculation of human marine fractions could result from observed slightly "marine" b13C sig- nals, which we have observed in Greenland reindeer bones (cf. Tauber 1984), unless this turns out to be due to consumption of seaweed. Clarification of these details will be gained by more extensive isotopic measurements on archaeological bone samples of the domestic animals and game consumed by the Norse.

The b13C in bone collagen mainly reflects the 13C composition of the protein in the diet (Ambrose and Non 1993), while bone apatite values may be more representative of the whole diet (see also Lanting and van der Plicht 1996). In the present study, the protein dietary composition may well be representative of the whole diet, as both the marine and terrestrial food resources of the Norse must have been very protein-rich. Since the Greenland climate did not allow cultivation of grain and veg- etables, the terrestrial diet of the Norse was limited to meat and dairy products. We therefore interpret the S13C variability in human bone collagen discussed in the following section as being directly representative of real differences in subsistence patterns.

NORSE CHRONOLOGY AND DIET

Five sample dates help us to establish the main chronological framework of the Norse in Greenland, even without relying on the reservoir correction procedure or archaeological dating. The ox bone from Brattahlid (AAR-1273) and the 3 burial textiles from Herj olfsnes (AAR-1288, -1289, -1290) establish the presence of Norse colonies in Greenland from around AD 1000 to the first half of the 15th century (AD 1430 ± 15), in other words slightly later than the last written historical evidence (a wedding in 1408). Arguments based on stylistic evidence from the "Burgundy Cap" for Norse presence after AD 1500 have been refuted by the 14C date (AAR-2201, Table 2) on the cap that places it in the 14th century (see Arneborg 1996).

The detailed chronology established by the 14C dates of Viking bones allows us to assess whether the large variation in marine food consumption may show a temporal pattern. The plot (Figure 3) of b13C values, the proxy for marine food fraction, versus calibrated 14C age reveals that the average diet of the Norse people changed from 20% marine to 80% marine during the approximately 500 years that the settlement lasted. This trend, if representative of the Greenland Norse population as a whole, shows a marked shift in Norse diet from the initial settlement period to the period of depop- ulation. The Norse colonists adapted to marine food resources, although at 80% not quite to the extent of contemporaneous Eskimos.

The marine element should derive first and foremost from seal according to the animal bone assem- blages found in middens in Greenland Norse settlements. Fish bones are nearly absent in the collections. This may, however, be explained by taphonomic biases such as the well-known poor preservation of fish bone and its appeal as a food source to both birds and domestic animals like dogs. It is hard to believe that the Greenland Norse did not tap the very rich resources of fish in the fjords, as did their relatives all over the North Atlantic and in Norway. Future isotope research may shed light on the extent to which fish actually formed part of the Greenland Norse diet.

The present study has provided a chronology and direct evidence for dramatic changes in the lives of the Norse in Greenland. The reason for the change may be deteriorating climate, which several authors have discussed as the main reason for the decline in population of the Norse settlements (e.g. 166 J Arneborg et al.

A Herjolfsnes Brattahli 0 GarBar Undir Hofa 100 ® Nonnen -13 Q Sandnes

-14 80

-15 0 0 -16 60

-17 D 40 -18

O Bishop -19 20 -20 Ox f I -21 - - - 0 900 1000 1100 1200 1300 1400 1500

Year AD

Figure 3 Measured 813C versus calendar year for Norse skeletons from Greenland. Symbols refer to churchyards. Sandnes is in the Western Settlement, the others are in the Eastern Settle- ment. Right-side ordinate axis is the marine food fraction calculated from 813C. The scatter in 813C is due to real differences in diet of individuals. Note the terrestrial value for the ox bone. High social status or recent arrival from Norway could explain the low marine diet of the bishop buried in the Gardar cathedral.

Stuiver et al. 1995; Barlow et al. 1997; Lynnerup 1998; see also Fricke et al. 1995). Recently, Green- land ice-sheet temperature data have been used for a climatic reconstruction, which is specific for the high-latitude North Atlantic region (Dahl-Jensen et al. 1998). The results indicate a steady decrease from a temperature maximum around AD 1000 to a minimum around AD 1500, which may have gradually forced the Greenland Norse to change their subsistence pattern.

ACKNOWLEDGMENTS

We acknowledge the skillful laboratory assistance of Vibeke Jensen (AMS Laboratory) in preparing the samples and graphite targets and also that of Guobjorg Aradottir (Science Institute) in operating the mass spectrometer for b13C measurements. We thank Egon Jans for his invaluable assistance in maintaining the ion sources and the tandem accelerator. Tinna Mobjerg has kindly let us use the 14C and b13C results measured by us on a series of Greenland Eskimo bones, submitted and funded by her as part of a separate research program. Change of Diet of the Greenland Vikings 167

The AMS Laboratory is financially supported by the Danish Natural Science Research Council. For 14C b13C the dates and measurements we gratefully acknowledge support from the following funds: Queen Margrethe and Prince Henrik's foundation, Regensburg's grant, the Augustinus Foundation, the Novo Nordic Foundation, Aase and Jorgen Minter's Foundation.

REFERENCES

Ambrose SH, Non L. 1993. Experimental evidence for Hansen FCC. 1924. Anthropologia medico-historica the relationship of the carbon isotope ratios of whole Groenlandiae antiquae. I. Herjolfsnes. Meddelelser diet and dietary protein to those of bone collagen and om Gronland 67:291-547. carbonate. In: Lambert JB, Grupe G, editors. Prehis- Hedges REM, van Klinken GJ.1992. A review of current toric Human Bone. Berlin: Springer-Verlag. p 1-37. approaches in the pretreatment of bone for radiocar- Arneborg J. 1990. The Roman Church in Norse Green- bon dating by AMS. Radiocarbon 34(3):279-91. land. Acta Archaeologica 61:142-50. Heier-Nielsen S, Heinemeier J, Nielsen HL, Rud N. Arneborg J. 1996. Burgunderhuer, baskere og dode nord- 1995. Recent reservoir ages for Danish fjords and ma- boer i Herjolfsmes, Gronland (Burgundian caps, rine waters. Radiocarbon 37(3):875-82. Basques and dead Norsemen at Herjolfsnes, Green- Heinemeier J, Rud N. 1997. Kulstof 14 datering med ac- land). English summary. Nationalmuseets Arbejds- celeratormetoden (AMS). Gronland 5-7:232-8. mark 1996. Copenhagen: Nationalmuseet. p 75-83. Hoegh OA. 1982. Tang I, Norge. Kulturhistorisk Leksi- Barlow LK, Sadler JP, Ogilvie AEJ, Buckland PC, Am- kon for Nordisk Middelalder 18:124-8. orosi T, Ingimundarson JH, Skidmore P, Dugmore AJ, Johansen OS, Gulliksen S, Nydal R. 1986. S13C and diet: McGovern TH. 1997. Interdisciplinary investigations analysis of Norwegian human skeletons. Radiocarbon of the end of the Norse Western Settlement in Green- 28:754-61. land. The Holocene 7:489-99. Jones G. 1986. The Norse Atlantic Saga. New York: Ox- Brown TA, Nelson DE, Vogel JS, Southon JR. 1988. Im- ford University Press. 337 p. proved collagen extraction by modified Longin Jonsson F. 1930. Det gamle Gr4nlands beskrivelse of method. Radiocarbon 30(2):171-7. Ivar Bkroarson. Copenhagen: Levin & Munksgaard. Broste K, Fischer-MOller K, Pedersen P0.1944. The me- 75 p. diaeval Norsemen at Gardar. Meddelelser om GrOn- Kieffer-Olsen J. 1993. Gray og gravskik i det middelal- land 89(3):1-62. derlige Danmark. Afdeling for Middelalder-arkceologi Chisholm BS. 1989. Variation in diet reconstructions og Middelalder-arkceologisk Nyhedsbrev (Aarhus based on stable carbon isotopic evidence. In: Price University):1-212. TD, editor. The Chemistry of Prehistoric Human Kristjansson L. 1982. Tang I, Island. Kulturhistorisk Lek- Bone. Cambridge: Cambridge University Press. p 10- sikon for Nordisk Middelalder 18:128-30. 37. Krogh KJ. 1967. Viking Greenland. Copenhagen: The Chisholm BS, Nelson DE, Schwarcz HP. 1982. Stable National Museum. 182 p. carbon isotope ratios as a measure of marine versus Lanting JN, Van der Plicht J. 1996. Wat hebben Floris V, terrestrial protein in ancient diets. Science 216:1131- skelet swifterbant S2 en visotters gemeen? [resume in 2. English]. Palaeohistoria 37/38:491-519. Chisholm BS, Nelson DE, Schwarcz HP. 1983. Marine Liden, K, Nelson DE. 1994. Stable-carbon isotopes as di- and terrestrial protein in prehistoric diets on the Brit- etary indicator in the Baltic area. Fornvannen 89:14- ish Columbia coast. Current Anthropology 24:396-8. 21. Dahl-Jensen D, Mosegaard K, Gundestrup N, Clow GD, Longin R. 1971. New method of collagen extraction for Johnsen SJ, Hansen AW, Balling N. 1998. Past tem- radiocarbon dating. Nature 230:241-2. peratures directly from the Greenland ice sheet. Sci- Lovell NC, Nelson DE, Schwarcz HP. 1986..Carbon iso- ence 282:268-71. tope ratios in paleodiet: lack of age or sex effect. Ar- Fischer-Moller K. 1942. The mediaeval Norse settle- chaeometry 28:51-5. ments in Greenland. Meddelelser om Gronland 89(2): Lynnerup N. 1998. The Greenland Norse - a biological- 1-82. anthropological study. Meddelelser om GrOnland - Fricke HC, O'Niel JR, Lynnerup N. 1995. Oxygen iso- Man & Society 24:1-149. tope composition of human tooth enamel from medi- Lynnerup N, Brings Jacobsen JC, Thorsen C, Kludt T. eval Greenland: linking climate and society. Geology 1997. Menneskeknoglerne fra gravene ved Asum- 23:869-72. miut. GrOnland 5-7:227-32. Grummesgaard-Nielsen S. 1997. Thulekulturens grave. Norlund P. 1924. Buried Norsemen at Herjolfsnes. Med- Gronland 5-7:198-227. delelser om Gronland 67:1-270. Gulliksen S, Scott M. 1995. Report of the TIRI work- Norlund P. 1930. Norse ruins at Gardar. Meddelelser om shop. Radiocarbon 37:820-1. Gronland 76:1-170. 168 J Arneborg et al.

Redin L. 1970. Lagmanshejdan: ett gravfelt som spegling historic man in Denmark. Nature 292:332-3. in relation av sociala strukturer i Skanor. Acta archaeologica Tauber H. 1984.14C Dating of human beings edi- Lundensia 10:1-201. to dietary habits. In: Mook WG, Waterbolk HT, Sympo- Renfrew C, Bahn P. 1991. Archaeology: Theory, methods tors. Proceedings of the First International 14C and practice. London: Thames and Hudson. 543 p. sium on and Archaeology. PACT 8. Strasbourg: Roussell, Aa.1936. Sandnes and the neighbouring farms. Council of Europe. p 365-75. photosynthe- Meddelelser om Gronland 88(2):1-219. Van der Merwe NJ. 1982. Carbon isotopes, Stuiver M, Braziunas TF. 1993. Modeling atmospheric sis, and archaeology. American Scientist 70:596-606. 13C 14C influences and 14C ages of marine samples to Van der Merwe NJ. 1989. Natural variation in con- 10,000 BC. Radiocarbon 35(1):137-90. centration and its effect on environmental reconstruc- DT, Stuiver M, Grootes PM, Braziunas T. 1995. The GISP2 tion using13CI12C ratios in animal bones. In: Price 8180 climate record of the past 16,500 years and the editor. The chemistry of prehistoric human bone. role of the sun, ocean, and volcanoes. Quaternary Re- Cambridge: Cambridge University Press. p 105-25. search 44:341-54. Van der Merwe NJ, Vogel JC.1978.13C content of human 14C Woodland Stuiver M, Polach HA. 1977. Discussion: reporting collagen as a measure of prehistoric diet in data. Radiocarbon 19(3):355-63. North America. Nature 276:815-6. Stuiver M, Pearson GW, Braziunas T. 1986. Radiocarbon Vebaek, CL. 1991. The church topography of the Eastern age calibration of marine samples back to 9000 cal yr Settlement and the excavation of the Benedictine Con- BP. Radiocarbon 28(2B):980-1021. vent Narsarsuaq in the Uunartoq Fjord. Meddelelser Stuiver M, Reimer PJ. 1993. Extended 14C data base and om Gronland - Man & Society 14:1-81. revised CALIB 3.014C age calibration program. Ra- Vogel JC, Fulls A, Ellis RP. 1978. The geographical dis- diocarbon 35:215-30. tribution of Kranz grass in South Africa. South African Tauber H. 1981. '3C evidence for dietary habits of pre- Journal of Science 74:209-15. RADIOCARBON, Vol 41, N r 2, 1999, p 169-182 ©1999 by the Arizona Board of Regents on behalf of the University of Arizona

USE OF RADIOCARBON DATING IN ASSESSING CHRISTIAN CONNECTIONS TO THE DEAD SEA SCROLLS

G A Rodley RD2 Picton, New Zealand

B E Thiering c/o School of Studies in Religion, University of Sydney, 2006 Australia

ABSTRACT. We present an analysis of radiocarbon dates on Dead Sea Scrolls that have a bearing on the question of the Scroll documents' relation to Christian origins. We assess details of dating reports, discuss paleographical evidence, and consider the content of the documents. When collated, these findings may be seen as compatible with a view that personalities mentioned in the Scrolls were contemporary with the founders of Christianity.

INTRODUCTION

Radiocarbon dating provides important information about the dates of some of the Dead Sea Scrolls, a matter of interest to the claim that they are closely connected with earliest Christianity.

The manuscripts properly called Dead Sea Scrolls were found at Qumran, near the northwest corner of the Dead Sea. Other, related documents that have been 14C dated were found at nearby locations in the Judean Desert. All historical indications point to their production in the general period of Jew- ish history extending from the third century BCE to the second century CE.

14C tests make it possible to give, within a statistically probable range, a date of manufacture of the writing material, either parchment (made from animal skin) or papyrus. This is not necessarily the same as the date of composition of the works inscribed on them. Some of the Dead Sea Scrolls, for example, were copies of the Old Testament, composed long before. Nor is the date of manufacture of the material necessarily exactly the same as the date when it was written on; manufacture may have preceded scribal use by a longer or shorter period of time. But in exceptional circumstances, an example of which will be given below, the date of manufacture can provide the dates of both composition and recording of the work.

The Dead Sea Scrolls that have aroused most interest are new works, not previously known, usually called the sectarian works. The contents of some of them give rise to the question whether the writers were in the immediate background of earliest Christianity, or were indeed part of that history themselves. Figures appear in these writings, always under pseudonyms, one called the Teacher of Righ- teousness, another referred to as the Wicked Priest. The dates when they lived, within the broad period of the Scrolls' production, are uncertain, and their identities have only been conjectured. The date of the writing material on which works discussing them are inscribed is, therefore, relevant to the question of their historical date, although not by itself conclusive.

Two 14C dating studies have been made, one by Bonani et al. (1991, 1992) and another by Jull et al. (1995, 1996). The Bonani et al. (Zurich) study included only 2 documents of direct relevance to the argument for a Christian connection, but because the whole study provided a good match between 14C values and paleographical assessments, this led to the view that the placement of the Teacher in the BCE period had been confirmed based on paleographical grounds. In the case of the Jull et al. Tucson study, comments made in the reports make it clear that the study aimed to determine whether the Scrolls have Christian connections, and it included 7 documents of direct relevance. The results of both studies have been examined by G Doudna of Copenhagen (Doudna 1998); the present study considers some of the questions he has raised.

169 170 G A Rodley and B E Thiering

14C DATING OF SCROLL SAMPLES

Extension of Results, Using Stuiver (1998) Calibration Dating of the Scrolls is complicated by the existence of 14C values, especially around 1900-2100 BP, that can correspond to more than one cal BCE/CE range. As a consequence, Scroll dates overall have somewhat wider 16 calibrated age ranges than might otherwise be expected for the level of accuracy obtained by the 14C measurements. All calibrated ages presented from the Tucson and Zurich studies were derived using the bidecadal Stuiver and Pearson (1986) calibration curve, except for an invocation of the decadal Stuiver and Becker (1986) curve in one instance by Jull et al. (1995). In Table 1 we use ranges derived from the Stuiver et al. (1998) curve (as reported by Doudna 1998) for the two Zurich and the seven Tucson documents of relevance, since these show some important differences from the Stuiver and Pearson (1986) ones. In some instances these contain separate subranges, due to the variations mentioned above. But these are of secondary significance for the dating considered here.

The most significant change is to the more recent extreme of the 1 6 range for a key document, 1QpHab. The relevance of this change is considered in more detail below.

Table 1 Stuiver et al. (1998) 16 calibrated age ranges (from Doudna 1998, Table A) with originally reported (Stuiver and Pearson 1986)16 values given in brackets Scroll Calibrated age range Zurich laboratory 11QTa (Temple Scroll) 53 BCE-21 CE BCE-1 CE] 1QH (Thanksgiving 37 BCE-68 CE [21 BCE-61 CE] Scroll) Tucson laboratory 4Q266 CE CE] 1 QpHab 88-2 BCE [104-43 BCE] 1QS 164-144 BCE; 116 BCE-50 CE [159 BCE-20 CE] 4Q258(second sample) 36 BCE-81 CE [11 BCE-78 CE] (4QpPsa) 4Q17 1 29-81 CE [22-78 CE] 4Q521 39 BCE-66 CE [35 BCE-59 CE] 4Q267 168 -51 BCE [172-98 BCE]

Features of the Dead Sea Scrolls Dating Studies Considerable care is required in the preparation of Scroll samples for radiocarbon analysis because of various historical storage conditions and chemical treatments of these documents (Caldararo 1995). Both groups who have made measurements addressed these problems carefully. Bonani et al. specifically checked the effect of gelatinization on three documents (two of which are ones cited QTa 14C here, l 1 and 1 QH). They concluded that gelatinization does not affect the age. The Tucson laboratory distinguished two categories of parchment sample, "Type 1", which appeared to be relatively clean, and "Type 2", contaminated with perspex glue. Reanalysis of one of the latter, 4Q258, suggested that an earlier measurement may have involved insufficient pretreatment of the sample. Of the seven Scroll samples studied by Jull et al. that are relevant to the question of Christian connections, only 4Q266 (in addition to 4Q258) fell into the Type 2 category; the others 14C Dating and the Dead Sea Scrolls 171 cited here were Type 1. The possibility of contamination by castor oil, used to highlight letters by the first generation of Scroll scholars, has also been raised; we consider this question below. A feature of the Zurich study was an apparent displacement of basic cal BCE/CE values toward older ages than expected from known dates and those estimated from paleographical studies, as noted by Bonani et al. (1992) and analyzed by Rodley (1993). However, as pointed out to us by Doudna (personal communication 1998; unreferenced), if the more recent calibrations are applied to the Zurich results, key features of the displacement disappear.

The Tucson study is linked to the Zurich one through their common measurement of one document, 1QIsa. The value determined by Tucson is in good agreement with Zurich. Consequently, we use Stuiver et al. (1998) derived ranges for the two Zurich documents related to the case presented here, the Temple Scroll and the Thanksgiving Scroll (Table 1).

14C Dating of Date-Bearing Papyrus Documents

In both the Zurich and Tucson studies, measurements were also made of certain documents as a check of the carbon-dating procedure. These documents bore an actual date, because they had the status of legal records. Al! of them were papyri. The cal BCE/CE values of Bonani et al. gave good agreement with the known dates, while the cal BCE/CE values of Jull et al. in all three cases showed a younger-age displacement from the known dates (Dull et al. 1995, Figure 1). In the case of one date-bearing sample, from Kefar Bebayou, the known date of writing, 135 CE, is well before the 16 Stuiver et al. (1998) range of 237-340 CE and even before the broader 26 range of 140-390 CE. Obviously, the papyrus was not cut after the date of recording, so some other common factor appears to have affected these measurements. Jull et al (1995, p 16) commented that the explanation might be that at the point involved (close to 130 CE) the calibration curve needed to be "slightly lowered". But the Zurich set of specific-age samples included two in just this age region, and the good agreement found for those would appear to eliminate the curve-adjustment possibility.

Instead, the displacement may indicate contamination, and a suggestion may be made about its source. (This would apply only to papyrus, made from reeds, not to the parchment, made from animal skin, on which the scrolls relevant to the Christian history were written.) The Tucson papyri 14C values may reflect insufficient acid pretreatment to eliminate all inorganic carbonate. The papyrus used would have grown in areas containing significant calcium carbonate (crystallization of the aragonite form is a notable feature of the region). Thus, it is possible that the papyrus writing material may have contained small crystals of aragonite. Such crystalline material would have contained 14C that was contemporary with the time of papyrus growth, in which case residual inorganic carbonate would not have affected the measurements. But the possibility exists that, over the historical existence of the papyri samples, more recent 14C became incorporated into the aragonite. That could occur via dynamic equilibrium of the aragonite carbonate with atmospheric carbon dioxide, facili- tated by the presence of moisture. Contemporary exposure to moisture could have enhanced such a process, thereby adding a significant amount of very recent 14C. If acid treatment had not been sufficient to remove all carbonate-and this becomes more likely for any embedded crystalline material-some "additional" inorganic 14C may have remained.

Acid treatment in the case of the Tucson study was carried out for a shorter time and at lower temperature than in the Zurich study. Although a somewhat higher acid concentration was used by Jull et al., the kinetic factors of time and temperature could have been the more significant ones with respect to complete removal of carbonate. Consequently, the residual presence of more recent inorganic 14C may have contributed to the age displacement observed for the papyri samples. This pos- 172 G A Rodley and B E Thiering sibility could be tested by remeasuring new samples of the papyri documents using more extensive acid pretreatment. As mentioned, carbonate in papyri samples would be embedded in the fibers as the result of formation during the growth of the papyrus. By contrast, Scroll parchments, made from animal skin, would be potentially subject only to surface effects capable of being readily removed by the cleaning/acid treatment.

CONTENTS OF THE SCROLLS AND THEIR BEARING ON DATING Along with the date of manufacture of the writing material, the contents of the documents provide evidence of the date of composition and recording of the sectarian scrolls. After describing the general principles for using such evidence, we apply them to specific documents. As in the case of the papyri, a document from this period may contain the actual date of its composition and recording. Let us suppose a hypothetical document A, bearing a date equivalent in our 14C terms to 1 March, 50 CE. But the date for the preparation of its writing material (corresponding to the killing of the animal whose skin was used, or the cutting of the papyrus), turns out to be mid-first century BCE. It would then be certain that the scribe of this document used an old piece of writing material. This is always possible in a situation where writing materials were scarce and valuable. Actual evidence for the use of old pieces will be given below.

Or, if a document contained no date, it might still describe certain events whose date is known to us. For example, a hypothetical document B might describe in exact detail the fall of Jerusalem, which took place in 70 CE. If its writing material were 14C-dated to the previous century, it would again be the case that the scribe used an old piece of material.

In a different and exceptional set of circumstances, the 14C dating could give positive information about the date of the events described, a date otherwise unknown. Let us suppose a hypothetical document C, which describes the activities of a person X whose date is not known to us. That person is said to be still alive at the time of writing. He must therefore have been alive after the date of manufacture of the writing material. The only exception to this rule would be a case where document C is a copy of an earlier document. But if it is certain that document C was the original and not a copy, then we have discovered that person X lived after the 14C date of document C. The record of his activities could not have been made on material that did not yet exist.

Key Documents for the Date of the Teacher

It is important to distinguish documents naming or directly reflecting the presence of the Teacher of Righteousness from other Qumran documents. The scribal community was at work over a long period, and some of their productions were composed early in their history. But one distinctive group having common characteristics, with similar organizational features and doctrinal emphasis, includes documents naming the Teacher of Righteousness as a great authority. Within this group, some name the Teacher of Righteousness as a great authority. It is this group whose date is significant.

Of the 7 pieces tested by Jull et al, that are relevant to the Christian question, 3 are from documents directly naming the Teacher of Righteousness: 4Q171, 1QpHab, and 4Q266 (see below on the related 4Q267). The Teacher was a figure who appeared at a certain point of the Qumran history. The works naming him uphold his authority against an opponent or opponents, also called by pseudonyms, and the conflict between them is believed to be a matter of such magnitude that it is claimed to have been predicted in Old Testament scripture. 14C Dating and the Dead Sea Scrolls 173

There are only indirect indications of the date of the Teacher, and the interpretation of these has been a matter of debate. 4QpPsa The documents called (4Q171) and 1 QpHab are examples of a genre of which a number of instances were found in the caves. Each is a pesher (plural, pesharim), that is, a commentary on a book of the Old Testament, claiming that its wording predicts events in the life of the Teacher and his opponents. Significantly, only one copy of each pesher was found in the caves, whereas multiple copies of other documents were found. The pesharim are an ephemeral genre of literature, referring to events in the author's own immediate circumstances, with the claim that they fulfilled prophecies. Once events changed, and the prophecies were seen to fit subsequent events better, the earlier document would be regarded as invalid. This would mean that no copies were made, each of the pesharim being an original.

(As the fact that there was only one edition of each of the pesharim is the foundation of this argument, a brief summary of the facts is given here. Eighteen fragmentary pesharim were found in the caves, 6 [3Q4, 4Q16!, 4Q162, 4Q163, 4Q164, 4Q165] commenting on different sections of Isaiah; 2 [4Q166, 4Q167] on different sections of Hosea; 2 on different sections of Micah [1Q14, 4Q1681; 1 on Nahum [4Q169]; 1 on Habakkuk [ 1 QpHab]; 2 on different sections of Zephaniah [1Q15, 4Q170];1 on Malachi [5Q10], 3 on different Psalms [4Q171, 1Q16, 4Q173]. Of the commentaries on Isaiah, all deal with different parts of the book, indicating that each pesher dealt with only a portion of it. In the one case where 2 different pesharim [4Q161, 4Q163] deal with the same 2 verses [Isa 10:20-22], it is plain that they are different compositions, for they break up the Isaiah verses dif- ferently when they add the pesher, and the pesher is different in each case). 40171(4QpPSa)

4Q171 is a pesher on some Psalms. At the time it was composed, the Teacher of Righteousness was still alive, and under threat from his opponents. The pesharist, his supporter, turned to Psalm 37 and found there the teaching that even though the righteous person may suffer now, he will soon be vindicated and his enemies punished. He applied this, using his technique of turning universals into par- ticulars, to the Teacher, who, he said, would soon be vindicated, while the opponents would be destroyed.

The principles of hypothetical document C apply to this work, the Teacher being person X. He was still alive at the time it was recorded, and the document we have, being a pesher, is an original, not a copy. These exceptional circumstances mean that the earlier date of the 14C range (Table 1) for the manufacture of the writing material gives us a probable date after which the Teacher was alive.

Both sets of 14C ranges, that based on the 1986 curve, and that derived from the 1998 curve (Table 1), give a date in the 1st century CE for this document: respectively, 22-78 CE and 29-81 CE. If it is the case that the 14C dating is a reliable indicator (as further argued in the section "Doudna's Argu- ments"), this result gives good evidence that the document was composed and recorded after the twenties CE, and, since the Teacher was alive and active at the time of composition, he lived in the first Christian century, being contemporary with the early Christians. It would be clear evidence against the view held by the first generation of Scrolls scholars, that the Teacher lived in the second century BCE.

The handwriting of this piece is a Herodian semiformal (Cross 1961, note 134; Strugnell 1970, p 211), a fact that is omitted in the Tucson report. The handwriting of all the pesharim is Herodian, that is, a class of handwriting used from 30 BCE to 70 CE. (Strugnell [1970] has treated 4QpIsac 174 G A Rodley and B E Thiering

on [4Q163], written on papyrus, as Hasmonean because it is in the same class as 1QS, but see below the paleography of 1QS. He has also raised the possibility that the "vulgar semiformal" hand of 4QpIsab (4Q162) is pre-Herodian, but such close distinctions in semiformals made by the early paleographers may now be doubted.) This finding agrees with one possible interpretation of the indirect datings given for the Teacher in the Damascus Document (CD). They are not overt, and their interpretation has been disputed, but when the usages of the Scrolls are applied consistently, they may be seen to mean that the Teacher began his work in 26 CE and died about 30 CE. The reasons in brief summary are as follows:

1. The wording of CD 1:5-11, concerning "the Period of Wrath, 390 years for his giving them into the hand of Nebuchadnezzar king of Babylon" is more correctly and consistently seen as a prediction of the length of the Roman occupation of Judea, the figure of 390 years being drawn from Ezekiel 4:5, treated as a prophecy in the habitual Qumran manner. In the pesharim it is asserted that "Babylon" of the Old Testament is an equivalent for Rome, a view found also in the New Testament, where "Babylon" is used as code for Rome (1 Pet 5:13, Rev 18). On this understanding, the ruler of Rome is being referred to by the writer of CD in a disguised way through a pseudonym, for political reasons. The usual translation of the phrase following "390 years" is "after his giving them", but it should be "for his giving them", consistently with the normal meaning of the preposition. The Roman occupation of Judea, an event that could well be called "the Period of Wrath", took place in 6 CE (see further below). Since, according to the text, the Teacher came 20 years after the Period of Wrath, he began working in 26 CE. 2. The writer of CD 20:13-15 calculates that the death of the Teacher occurred about 40 years before a certain destruction of enemies. The destruction would result from a Visitation, an event expected in the near future. The Visitation is described in CD 19:10-16 using the language of the first fall of Jerusalem (Ezek 9:4). The passage may be understood as referring to another fall of Jerusalem, very shortly expected at the time of the writer. Jerusalem did fall in 70 CE; hence, on this interpretation, the Teacher died about 30 CE. (See Thiering 1979).

1 QpHab: More Accurate 16 Range 1QpHab, an extensively preserved pesher, is a commentary on the Old Testament book of Habak- kuk, with frequent references to the Teacher and his rivals. He is spoken of as a past figure who had been defeated by his rivals, so it was composed after 4QpPsa. Events in which he had been involved were vividly present to the writer, so, given the ephemeral nature of the pesharim, it would not have been composed too long after the events. The first published 14C date for this document was 104-43 BCE, based on the 1986 curve. But the 1998 curve gives a date of 88-2 BCE (Table 1).

The handwriting of 1 QpHab is Herodian. It was copied by the same scribe as the second copy of the Temple Scroll, in a hand described as a developed Herodian formal from about 20-50 CE. (DJD 23, p364). Given these facts, and given the first century CE date of 4QpPsa, which preceded this document, 4QpPsa there is no great difficulty in supposing that a writer coming shortly after recorded 1 QpHab on a piece of writing material that was somewhat older than the writing material used by his col- league. The question of the use of an older piece of material may be discussed in conjunction with the reaction of some Scrolls scholars to the Tucson date for this document, which they took to be definitive, and to be evidence against the Christian connection. In a footnote to their 'Atigot publication of the '4C Dating and the Dead Sea Scrolls 175

results Jull et al. (1996) quoted a letter of 29 November 1992, from G. Vermes, a translator of the Scrolls, recommending that 1 QpHab be tested: "If the carbon dating establishes (for 1 QpHab) a ter- minus ad quern prior to ca. 30 CE, this will damage almost beyond repair the hypothesis proposing a Christian connection".

But even in the absence of the subsequent calibration curves that render the date later, this was a historical non sequitur. Although a work could not be recorded on material that was not yet manufactured, it could well be recorded on material that had been manufactured many years before. Two pieces of evidence indicate that old materials were used at Qumran.

The first is a statement by Josephus that the Essenes (who formed the nucleus of the Qumran sectarians, even though ascetics of other outlooks joined them) "do not change their garments or shoes until they are torn to shreds or worn threadbare with age" (Jewish War 2, 126). This indicates that they had no objection to using old materials, and some may have found religious or ascetic reasons for doing so. Parchment remained usable for a long time, as is shown by the condition of pieces found in the caves after 2000 years. The parchment, "although hard and brittle when received, became relaxed and flexible on exposure to moisture" (DJD 1, p 39). The other possible piece of evidence is the second sample in the Zurich list, the Testament of Qahat, 4Q542, whose 16 14C date was given as 388-353 BCE or 309-234 BCE, but whose handwriting is paleographically dated to 100-75 BCE. (PAM 42.600, Eisenmann and Robinson 1991, plate 923; also PAM 43.565, Eisenmann and Robinson 1991, plate 1513). It uses letter forms found in the middle or late Hasmonean period, and it is certainly not an archaic script such as would match its 14C dating. The authors of the Zurich report judged that "the possibility that the leather was preserved uninscribed for such a long period is quite unlikely" (although without taking into account the Ess- ene habits), and suggested that contamination in this specific case could not be ruled out. Any contamination must have come from an even older source, since contamination by subsequently applied agents containing modern carbon gives a younger, not an older date than the true one. Pending further laboratory work, there is at least a possibility that a very old piece of material was used. In this connection, Doudna's discussion of the 14C dating of a linen wrapper that was probably found in Cave 4 is another example of not taking into account historical evidence for the use of old materials. He states that "the true date of the Cave 4 linen item is presumably close to the date when the scroll it was wrapped around was deposited in the cave", and goes on to suggest that this may be the date when all scrolls were deposited in Cave 4. The calibrated date range of the linen on the 1998 curve is 165-144 BCE or 117-2 BCE, 16. (The 26 date of 197 BCE-46 CE is not quoted in the argument). However, it has frequently been suggested by Scrolls scholars that Cave 4 was a Genizah, a burial place for "dead" scrolls. Jewish tradition held that any document containing the name of God must not be destroyed, but buried, with a funeral service, as if it were human. It would be an expression of reverence to wrap the "body" in linen that had been preserved for a long time.

Moreover, the contents of 1 QpHab may be seen as giving information about its date of composition. The situation of hypothetical document B applies to this work, which may be seen as describing events occurring in the first century CE. An army called the "Kittim" is described as a terrible and destructive force, marching across the land. It has long been accepted that the Kittim were the Romans, following Yadin's observations concerning the weaponry and military tactics of the Kittim, which were the same as distinctive practices of the Romans (Jeremias 1963). This is accepted by the most conservative of scholars, including Vermes (1995).

The Romans appeared in Judea as a terrible and destructive force only after 6 CE, when they occupied the country and put it under direct Roman rule. Prior to this, they had been distant overlords. 176 G A Rodley and B E Thiering

Pompey had indeed, in 63 BCE, brought the country under Roman dominion, but he is not presented as a malevolent, overwhelmingly destructive power, as are the Kittim of 1 QpHab. (The Kittim "march over the plain, destroying and plundering the cities of the earth.... The fear and dread of them is upon all the nations ... in a council are all their plottings for evil, and with cunning and deceit they deal with all the peoples ... they trample the earth with their horses and beasts. From afar they come, from the islands of the sea, to devour all the peoples like an insatiable eagle... . They sacrifice to their standards and worship their weapons of war", 1QpHab 3:1-12, 6:3-4). Although Pompey entered the Holy of Holies, he treated it with respect, earning Josephus' praise for his "virtuous character" (Antiquities 14, 72-73). He restored a Jewish high priest to his place, and the country was thereafter governed by its own native high priests and kings. Only in 6 CE, after the dismissal of Archelaus Herod, did Judea become an occupied nation under the direct government of Roman procurators. The outrage at their presence was so strongly felt that a band of militants was formed, later called Zealots. They harassed the Romans until their activities brought about the destruction of Jerusalem later in the century. Furthermore, there was a particular occasion in late 37 CE, recorded by Josephus, of a march of Roman soldiers across the land, when the army of the governor Vitellius passed through southern Judea, not far from Qumran. His action led to protest: "since he had started to lead his army through the land of Judea, the Jews of the highest standing went to meet him and entreated him not to march through their land. For, they said, it was contrary to their tradition to allow images, of which there were many attached to the military standards, to be brought upon their soil" (Antiquities 18, 120- 122). Vitellius yielded to their request, having learned from the fate of the recently dismissed Pon- tius Pilate, who had been the most oppressive of the procurators up to that time. But the appearance of Vitellius' 2 legions of heavy-armed infantry and auxiliary light-armed infantry and cavalry would have evoked an initial reaction, bringing to the surface the anguish felt since the occupation. An involvement of the Qumran sectarians in zealotry may be argued from the contents of their document, the War Scroll, and from the fact that fragments of scrolls were found at Masada.

These circumstances may be understood as the background of 1 QpHab, which would have been composed toward the end of 37 CE, at the time of the initial reaction to the march of Vitellius. The writer not only expressed his fear of the Romans, but looked back to events in the life of the Teacher not many years before, finding them all "predicted" in the book of Habakkuk. This is consistent with a date for the death of the Teacher in about 30 CE. 4QpPsa Doudna's Arguments Concerning and 1 QpHab While making valuable observations about the uncertainties of 14C dating, Doudna's discussion of 4QpPsa the two most significant documents for the Christian connection, and 1 QpHab, rests on unwarranted assumptions that are contrary to historical evidence or to the contents of the Scrolls. His main assumption, used not only in this case but in his treatment of other documents (4QTQahat and 4QLevia ar: Doudna 1998, p 445) is that if 2 documents, recorded by different scribes, are closely related in subject matter, then their writing materials must have been manufactured at similar dates. This is an unsustainable assumption, on general grounds. If two authors both discuss the same contemporary events, it does not prove that their writing paper was manufactured at the same time, or even in the same generation. Further, the evidence for the Essene use of old materials for religious reasons rules out this assumption altogether. Doudna, agreeing that the subject matters and styles of the two key documents are closely related, and that there was only one copy of each of the pesharim, says, "There is thus good reason to expect 4QpPsa that and 1 QpHab should be contemporaneous both in composition and in their single scribal 14C Dating and the Dead Sea Scrolls 177

copies. If this analysis is correct, the apparent difference in radiocarbon dates for 1 QpHab and 4QpPsa may represent not a real difference in dates, but rather an anomaly in the radiocarbon measurements" (1998, p 453). Later he speaks of "the older radiocarbon date for 1 QpHab, with which the scribal copy 4QpPsa ought to be contemporaneous" (p 461).

Choosing between the 2 dates, he concludes that 1 QpHab is the preferable one. He holds that 4QpPsa may be an "outlier", a measurement that differs from that of other similar items without known cause, possibly through error. But his reason for this comes from a hypothesis that is not rec- oncilable with the contents: that all the Qumran Scrolls belong together in a single generation. He says that "since 4QpPsa has the youngest radiocarbon date for Qumran texts in either laboratory's group, its results are a priori of less secure confidence than the dates for the others" (p 461).

Few Scrolls scholars would agree that all the Scrolls belong together in a single generation. The foundation studies of such documents as the Community Rule and the Damascus Document, and of their relation to other documents, showed a process of development of the community organization over time. Further, the documents naming or reflecting the Teacher are a special group, which must be treated separately, as noted above. Some of the Scrolls show no knowledge of the Teacher and his distinctive doctrines and organization. It is the pesharim, together with the Damascus Document, the Community Rule probably, the Hymns of Thanksgiving, and other possible inclusions, that are relevant to the question of the Teacher. Within this group, all of the documents 14C-dated so far are capable of placing him in the first Christian century, as will be further shown below. Of the pesharim, only 2 of the 18 have so far been 14C-dated. It is erroneous to compare 4QpPsa with documents to which it is not related, and to say on this basis that it is an "outlier".

In comparing 1QpHab and 4QpPsa on the question of contamination, Doudna gives two reasons why the former was not contaminated. Of these, the first reason-that it was never in the Rocke- feller Museum, where it is known that castor oil was used to make the letters clearer-is unpersua- sive, for if the use of castor oil was a routine procedure, as stated, it may have been employed elsewhere. The second reason is more likely, that the tested piece was from a large amount of blank space, which would not have been subjected to castor oil. But this point applies also to 4QpPsa, which has wide margins. Doudna, suggesting that 4QpPsa might have been contaminated, states that it was not subjected to acetone treatment, which would have removed castor oil, but he does not give any positive evidence that it was affected by castor oil. In fact, in his footnote 59 he allows that it is "intrinsically unlikely" that any given sample was so affected. (He does not mention that it was one of Tucson's "relatively clean" Type 1 parchment samples.)

It would appear that Doudna's final proposal, that "the first century CE disappears from Qumran's textual horizon" (1998, p 464) is not justified, either by his discussion of 14C dating, concerning which he has mainly emphasized its uncertainty, or by 14C dating taken in conjunction with paleography and the contents of the Scrolls.

40266: Evidence of a Particular Paleographical Error

When the Zurich results were published, it was apparent that they were in good agreement with the paleographical datings already established for the 14 documents tested. This fact was seized upon by some Scrolls scholars as evidence against a Christian connection, for it was on the basis of paleographical findings that the Teacher had been placed in the 2nd or 1st centuries BCE, and it was believed that paleography had been vindicated.

But such a conclusion had not looked closely at the detail of the paleographical findings. In fact, only 2 published documents appeared to put the Teacher so early in terms of paleographic dating. 178 G A Rodley and B E Thiering

They were 4Q266 and 1QS, both now 14C dated. 4Q266 is a copy of the work called the Damascus Document, which is not a pesher, but names the Teacher. 1QS, the Community Rule, reflects the Teacher's doctrine, although it does not name him.

On the rules of paleography alone, it was possible to demonstrate problems with the paleographical finding in both of these cases (Thiering 1979). As noted above, all the pesharim are in Herodian formal or semiformal scripts. If it were not for these two documents, 4Q266 and 1 QS, the Teacher could have been placed in the Herodian period (30 BCE-70 CE). But even one early document naming him, if it were certainly early, would be enough to place him before its time, in the Hasmonean period (150-30 BCE).

The script of 4Q266 had been announced to be early, and to place the Teacher somewhere before 60 BCE (Milik 1959, p 58). But the hand of this document is a semicursive, and the cursive and semicursive scripts are much more difficult to date than the formals and semiformals, as paleographers admit (Cross 1961, p 146, 182). They correspond to personal handwriting, whereas the formals were under controls like those of print, which are relied upon by paleographers. The writing of 4Q266 is described in the official publication (DJD 18) as a "rapid and careless hand". Doudna remarks that the 14C dating of 4Q266 (la 4-82 CE, 26 44 BCE-129 CE [Stuiver et al. 1998 values]) raises questions about the paleographical judgment. "It might be suggested that the high- precision paleographic estimates that have been given to these two texts (4Q266 and 4Q521) are somewhat premature" (1998, p 460). He leaves undetermined whether the radiocarbon evidence should be allowed to correct the paleographical conclusion. It may be argued however, that the discrepancy comes from giving too firm a date to a semicursive, a notoriously slippery class for paleographers. If the 14C dating is a reliable indicator, the 16 range, 4-82 CE, allows the inference from this document, taken alone, that the Teacher lived in the 1st century CE. However, the Damascus Document was found in multiple copies, and further factors must be taken into account in determining its date of composition. The 14C result for 4Q267 is relevant to it; see below.

40267: An Early Source of the Damascus Document Also tested (Tucson) was 4Q267, a more recently published fragment of the Damascus Document. The nature of the Damascus Document is relevant to the consideration of this piece and 4Q266. The complete version of this work is known to us from medieval copies that were found in Cairo and published in 1910. Many fragments of it were subsequently found in the Qumran caves, and it could be seen that it had originally come from a stage when Qumran sectarians were in Damascus. Although most of it is concerned with laws, parts of it deal with the history of the Teacher of Righ- teousness and his rivals, treating them in much the same way as does 1 QpHab. It appears that the rivals, who had become powerful in Judea, had been the cause of the sectarians' exile to Damascus. Whereas the fragments of 4Q266 represent a fairly extensive version of the Damascus Document, running parallel to the medieval copy, 4Q267 consists of only a small group of fragments. They are in an early Herodian formal hand (DJD 18, p 96). None of the extant parts of the fragments contain references to the history of the Teacher of Righteousness and his rivals. (A reconstruction of fragment 2, line 15, may give the phrase "law-interpreter", but according to 1 QS 6:6 this means a levite subordinate to a priest, whereas the Teacher was a priest. A reconstruction of fragment 3, line 7, may perhaps speak of "a teacher", but not of the Teacher of Righteousness. In both cases, the full version of the text, CD itself, goes on to speak of the Teacher of Righteousness, but if our argument that '4C Dating and the Dead Sea Scrolls 179

4Q267 is a source is correct, it will have been the case that the full version added these references to the source.)

Fragment 2 of 4Q267, although not referring to the Teacher, is relevant to the question, as it cone- sponds to CD 5:17b-6:7. In the full CD, 5:17b-6:7 is a new section immediately following an account of the Teacher and his disputes. This section, placed in column 5, differs from what pre- cedes, in that it introduces the word "Damascus" for the first time, in the context of the exile of the men of the community.

The 1998 calibration gives to 4Q267 a 16 date of 168-51 BCE, 26198-3 BCE. Following the argument above, that 4Q266 was recorded in the 1st century CE, with the implication that it may have been composed then, it would not be impossible to argue that 4Q267 was simply the remains of another copy of the work on an older piece of material.

However, a secondary question arises concerning CD itself, one that would make the argument for an old piece of material unnecessary. A close study raises the probability that 4Q267 is a source, composed in the first century BCE, which was subsequently incorporated into CD, a document that is known to have been composite. The stages will have been:

1. 4Q267, dealing with an exile to Damascus in the 1st century BCE, and justifying it as directed by God. 2. 4Q266, dealing with a subsequent exile to Damascus (a natural place for political exiles, just outside the boundaries of Judea) in the 1st century CE, the document now beginning with the recent history of the Teacher and his rivals. It incorporated parts of 4Q267, placing the Dam- ascus passage in its column 5, after its treatment of the Teacher. The inclusion was in order to endorse the argument concerning Damascus. 3. The final version, which was copied in medieval times.

The argument for this is as follows:

A. The wording of fragment 1 of 4Q267 corresponds, as is recognized, to the wording of another fragment of CD. This latter piece was found still attached to its fastening, proving that the words in question were part of an opening column. The piece attached to the fastening is in a semicursive script, whereas 4Q267 is in a formal script. But 4Q266 is in a semicursive script, and for that reason the piece attached to the fastening was placed with 4Q266, as its opening column.

But close study shows that it was an error to put the two semicursive pieces together. They are in a different semicursive hand. Although the fragment attached to the fastening was small, the following differences in the method of drawing the letters may be observed, showing a recurring characteristic, that the scribe of 4Q266 makes fuller strokes than does the other scribe:

Letter 4Q266 Fragment attached to fastening Lamedh full hook short hook Ayin full right arm short right arm Mem base and oblique nearly joined space between base and oblique Taw full left stroke short left stroke Beth long base more even top and base Shin long center and right strokes short center and right strokes Qof flat top, angle at right corner more rounded top Tet slanted base straight base 180 GA Rodley and B E Thiering

B. The wording of the piece attached to the fastening (which on this argument remains without a siglum), and of its parallel 4Q267 fragment 1, give further reasons for believing that they represent a document that was different from the full version of CD: 1) if the reading "sons of light" in line 1 of the attached fragment is correct, then the piece does not belong to the longer CD, where this significant term-often used as a basis for analysis-is not found. 2) The attached fragment uses the first person plural for the author in line 19. The first person plural is not used in the longer CD (except in an acknowledged quotation, 20:29).

C. These observations raise the probability that 4Q267 and the fragment attached to the fastening represent two different copies, in different handwriting, of the same document, a document that was not the same as the fuller CD. Parts of it were incorporated in the fuller CD, but not its opening column. Since the extant fragments of 4Q267 do not deal with the Teacher, it may reasonably be supposed that the document was a source preceding the time of the Teacher. Its early 14C dating, then, is not evidence for the early date of the Teacher.

Thus the inference from the 14C dating of 4Q266, that the Teacher lived in the 1st century CE (see previous section) is unaffected by the 14C dating of 4Q267.

1 oS and 40258: The Community Rule

1QS and 4Q258 are 2 copies of the same work, the Community Rule or Manual of Discipline. Its principal version is known to us in 11 complete columns, called 1 QS because it was found in Cave 1. The other, 4Q258, found in Cave 4, is a fragment, corresponding to 1QS 5:1 ff, in a different hand from 1QS. The work is in the form of regulations governing the life of the community, which was bound to obedience. Parts of it may be understood as containing doctrine and legislation like that of the Teacher, although he is not named. Because of characteristics of the 1998 calibration curve and the large error associated with the 14C measurement, the full document 1QS yielded two 16 ranges, 164-144 BCE and 116 BCE-50 CE, spanning a wide age range (Stuiver et al. 1998 values, Table 1).

The hand of 1QS is not a normal one, but an unusual, highly embellished one, said from the first to be atypical, and presenting a complicated problem (Cross 1961, note 116). Cross put it in the class of Hasmonean semiformal, making it early first century BCE, but it can be shown that it combines Herodian forms with letter forms used in the related Palmyrene scripts (Thiering 1979). Palmyrene scripts using the same forms are actually dated in the early Herodian period, and the forms continued to be used well into the Christian period.

The related sample, 4Q258, had to be tested twice because of possible contamination, but the second test yielded a 1c calibrated age within the Christian period, 36 BCE-81 CE (Stuiver et al. 1998 values, Table 1). Its hand has been described by Cross (1994) as "early Herodian formal". (This cor- rects the statement on its paleography in Jull et al 1995, p 18.) These facts are consistent with a view that the Community Rule was compiled during the first centuries BCE and CE as a legal document by which members of the Qumran community were bound. It is known to reflect different stages of organization. The argument would be that its final form, 1 QS, incorporating the Teacher's doctrine, was recorded in the 1st century CE in an embellished, unusual hand, to give it prestige, as legal documents still are. The 14C 16 range of 116 BCE-50 CE (Table 1) would permit a first century CE date of manufacture of the material (but see further below on 4QSamc). 4Q258, corresponding to part of it, would be a source that appeared during the process of compilation. 14C Dating and the Dead Sea Scrolls 181

Any discussion of 1 QS must take into account a document tested by the Zurich laboratory, 4QSamc, which was written by the same scribe as 1 QS (Cross 1961, note 116). The hand, as shown above, is an idiosyncratic one, combining late Palmyrene with Herodian features. The 14C 16 range of 4QSamc is 196-47 BCE (Doudna 1998, Table A). Since a first century CE date for the final form of 1 QS may be argued, as shown above, it would follow that the scribe of both works chose to use for 4QSamc a piece of some antiquity, consistently with Essene habits. This may also have been the case 14C with the piece he chose to record 1 QS, whose dating presents a very broad range. Given the embellished nature of the handwriting, and the authority of the subject matter in both cases, it is fea- sible that venerable pieces of parchment were used by this scribe.

40521

Another document in the Tucson group, 40521, called The Messianic Apocalypse, has been given a dating that could help to challenge a view that it was a very early composition. This work contains doctrinal parallels to the New Testament. It has aroused considerable interest for that reason, but it is less relevant to the historical question, as it does not deal with history or mention the Teacher. However, its placement in the late first century BCE or first century CE (16 calibrated age 39 BCE- 66 CE, Stuiver et al. 1998 value, Table 1) gives chronological plausibility to an argument for a historical connection between Qumran and the early Christians.

Its hand has been defined as Hasmonean formal (DJD 25, p 3). Making the usual allowance for "the extension of the professional life of a conservative scribe beyond his generation, or for an individu- alistic hand" (Cross 1961, note 29), it could have been used in the early Herodian period, indicating that its doctrinal concerns preceded the similar ones of Christians. In his official publication of the document (DJD 25, 1998) Puech refers to the 14C dating but cites only the broad 26 range, 93 BCE- 80 CE (Tucson, using the 1986 calibration curve).

The Two Zurich Documents Relevant to the History

The conclusion that the Zurich results gave evidence against the Christian connection came not only from inadequate observation of the paleographical detail, as shown above, but also from inadequate observation of the contents of the documents tested. Only two of them were relevant to the history. One was the Temple Scroll (11Q19, 11 QT'), which clearly came from a first stage of the sectarian development, preceding but related to subsequent works such as the War Scroll and the Community Rule. The other was the Hymns of Thanksgiving (1 QH), which was established by Jeremias (1963), on the basis of word frequencies and common terms, to have been partly composed by the Teacher of Righteousness.

It had been argued (Thiering 1979), from the contents of the documents alone, that the Temple Scroll was composed about 21 BCE, at the time Herod the Great announced his plan for rebuilding the temple. The Qumran sectarians, encouraged by Herod's approval of the Essenes at that time (as recorded by Josephus) were offering their plan for the new temple, claiming that it had been divinely revealed. This interpretation is entirely compatible with both the original Zurich dating (97 BCE-1 CE) and with the more recent calibration, 53 BCE-21 CE (Table 1). It is also compatible with the scroll's Herodian hand.

The Thanksgiving Scroll's calibrated age of 21 BCE-61 CE, given by Zurich, and 37 BCE-68 CE from the 1998 calibration, is also consistent with parts of it having been composed by the Teacher of Righteousness between 26 and about 30 CE. 182 G A Rodley and B E Thiering

CONCLUSION The radiocarbon dating of the relevant group of sectarian documents within the Dead Sea Scrolls, when correlated with paleography and the contents of the documents, allow the possibility that the Teacher of Righteousness lived and died during the period of the foundation of Christianity. This conclusion is contrary to a view now commonly held, that the radiocarbon datings have disproved a Christian connection.

ACKNOWLEDGMENT

We thank Drs A J T Jull and T Higham for their helpful assistance.

REFERENCES [DJD 1] Barthelemy D, Milik JT. 1955. Qumran Cave I. edition of the Dead Sea Scrolls. Washington DC: Bib- Discoveries in the Judaean Desert 1.Oxford: Claren- lical Archaeology Society. don Press. 165 p. Garcia Martinez E 1994. The Dead Sea Scrolls trans- [DJD 18] Baumgarten JM. 1996 Qumran Cave 4, XIII: lated: the Qumran texts in English. Leiden, Brill. 513 The Damascus Document (4Q266-273). Discoveries p in the Judaean Desert 18. Oxford, Clarendon Press. Jeremias G. 1963. Der Lehrer der Gerechtigkeit. Gottin- 236 p. gen, Vandenhoeck & Ruprecht. 376 p. [DJD 23] by Garcia Martinez F, Tigchelaar EJC, van der Jull AJT, Donahue DJ, Broshi M, Toy E. 1995. Radiocar- Woude AS. 1998. Qumran Cave 11, II: 11Q2-]8, bon dating of scrolls and linen fragments from the 11Q20-31. Discoveries in the Judaean Desert 23. Ox- Judean desert. Radiocarbon 37(1):11-19. ford, Clarendon Press. 487 p. Jull AJT, Donahue DJ, Broshi M, Toy E. 1996 Radiocar- [DJD 25] Puech E. 1998. Qumran Grotte 4, XVIII: Textes bon dating of scrolls and linen fragments from the hebreux (4Q521-4Q528, 4Q576-4Q579). Discover- Judean desert. 'Atigot 28:85-91. ies in the Judaean Desert 25. Oxford, Clarendon Press. Milik JT. 1959. Ten years of discovery in the wilderness 229 p. of Judaea London: SCM.160 p. Bonani G, Broshi M, Carmi I, Ivy S, Strugnell J, Wolfi, Rodley GA. 1993. An assessment of the radiocarbon dat- W. 1991. Radiocarbon dating of the Dead Sea Scrolls. ing of the Dead Sea Scrolls. Radiocarbon 35(2):335- 'Atigot 20:27-32. 8. Bonani G, Ivy S, Wolfi W, Broshi M, Carmi I, Strugnell Strugnell J.1970. Notes en marge du volume V des "Dis- J. 1992. Radiocarbon dating of fourteen Dead Sea coveries in the Judaean Desert of Jordan". Revue de Scrolls. Radiocarbon 34(3):843-9. Qumran 7(26):163-76. Caldararo N. 1995. Storage conditions and physical treat- Stuiver M, Becker B. 1986. High-precision decadal cali- ments relating to the dating of the Dead Sea Scrolls. bration of the radiocarbon time scale, AD 1950-2500 Radiocarbon 37(1):21-32. BC. Radiocarbon 28(2B):863-910. Cross, FM. 1961. The Development of the Jewish scripts. Stuiver M, Pearson GW. 1986. High-precision calibra- In: Wright GE, editor. The Bible and the ancient Near tion of the radiocarbon time scale, AD 1950-500 BC. East. London: Routledge & Kegan Paul. p 133-202. Radiocarbon 28(2B):805-38. Cross, FM. 1994. Paleographical dates of the manu- Stuiver M, Reimer PJ, Bard E, Beck JW, Burr GS, scripts. In: Charlesworth JH, editor. The Dead Sea Hughen KA, Kromer B, McCormac G, van der Plicht Scrolls, Hebrew, Aramaic and Greek texts with En- J, Spurk M. 1998. INTCAL98 radiocarbon age cali- glish translations. Volume 1. Tubingen: J. C. B. Mohr. bration, 24,000-0 cal BP. Radiocarbon 40(3):1041- p57. 1083. Doudna G. 1998. Dating the Scrolls on the basis of radio- Thiering BE. 1979. Redating the Teacher of Righteous- carbon analysis. In: Flint PW, Vanderkam JC, editors. ness. Sydney: Theological Explorations. 234 p. The Dead Sea Scrolls after fifty years. Volume 1. Vermes G. 1995. The Dead Sea Scrolls in English. Rev Leiden: Brill. p 430-71. and ext 4th ed. London: Penguin Books. 392 p. Eisenmann RH, Robinson JM, editors. 1991. A facsimile RADIOCARBON, 2, Vol 41, Nr 1999, p 183-197 ©1999 by the Arizona Board of Regents on behalf of the University of Arizona

14C AMS DATING OF EQUIPMENT FROM THE ICEMAN AND OF SPRUCE LOGS FROM THE PREHISTORIC SALT MINES OF HALLSTATT

Werner Rom Robin Golser Walter Kutschera Alfred Priller Peter Steier Eva M Wild Vienna Environmental Research Accelerator, Institut fur Radiumforschung and Kernphysik, Universitat Wien, Wahringer Strasse 17, A-1090 Wien, Austria

ABSTRACT. This paper summarizes radiocarbon measurements of mainly botanical samples from the Iceman ("Otzi") and from his discovery site, an Alpine glacier at the Austrian-Italian border. The results were obtained by accelerator mass spectrometry (AMS) at 3 different laboratories (Vienna, Austria; Uppsala, Sweden; Gif sur-Yvette, France) between 1992 and 1997. All the dates, except 2, are consistent with the time period 3360-3100 BC, as previously determined from bone and tissue samples from the Iceman himself. The 2 exceptional dates from wooden artifacts suggest that the site of the Iceman was used as a mountain pass for millennia prior to and after the lifetime of "Otzi".

For a 2nd sample complex, we studied logs from the beginning of salt mining in the world's oldest salt mines at Hallstatt in '4C Upper Austria. AMS measurements were performed in Vienna on spruce samples found in the prehistoric mines and from a log-house on the surface. Data evaluation included "wiggle matching" of different sets of tree rings. The results suggest that salt mining in the Hallstatt region took place in the l4th-l3th century BC, well before the so-called Hallstatt period.

We discuss in some detail the chemical pretreatment of the samples and the data evaluation. We also present a comprehensive 14C survey of dates available in the literature concerning both botanical remains from the vicinity of the Iceman and from the earliest salt mining in Hallstatt.

INTRODUCTION

Recently, 2 sample complexes of considerable archaeological significance to the early history of the Central European region were radiocarbon dated at the Vienna Environmental Research Accelerator (VERA).

Equipment from the Iceman

On 19 September 1991, a mummified corpse was discovered accidentally by tourists at the "Tisen- joch", a high Alpine mountain pass near the Italian-Austrian border (3120 m asl). The find was quickly nicknamed "The Iceman" or "Otzi" (after the site location in the Otztal Alps) and received widespread attention (see e.g. Coghlan 1992; Jaroff 1992; Bahn and Everett 1993; Roberts et al. 1993; Barfield 1994).14C accelerator mass spectrometry (AMS) measurements performed on bone and on tissue samples (Bonani et al. 1992, 1994; Hedges et al. 1992; Prinoth-Fornwagner and Niklaus 1994) proved that the Iceman is world's oldest known intact mummy. Many pieces of the Iceman's equipment and other materials associated with his location were recovered in several post-excavations (Lip- pert 1992; Spindler 1993; Bagolini et al. 1995), taking place shortly after and during the cleansing of the mummy at the Romisch-Germanisches Zentralmuseum in Mainz, Germany (Egg 1992). How- ever, only a very small fraction of these additional findings has been dated and published so far (Bonani et al. 1994; Prinoth-Fornwagner and Niklaus 1994). Botanical details of findings in connection with the Iceman are discussed in Bortenschlager et al. (1992) and Oeggl (1995). Here, we summarize the 14C dating measurements performed at 3 different AMS laboratories on a representative fraction of these mainly botanical samples (see also Kutschera et al. 1998).

Spruce Logs from the Prehistoric Salt Mines of Hallstatt

The salt mines of Hallstatt, the oldest in the world (Barth 1993; Lippert 1985), are probably the most important prehistoric industrial sites in Europe. They provided the basis for the enormous wealth of this region at that time. They lasted through 3 prehistoric periods (the (Late) Bronze Age, the Early Iron Age, and the period around 1 AD, respectively), which are topographically associated with the

183 184 W Rom et al. so-called Northern, Eastern and Western mining groups (Schauberger 1960; Lippert 1985). Each of these groups had its distinct salt production technology (Urban 1989; Barth 1993). The importance of salt for this region is reflected in the syllable "Hall", which is derived from the ancient Greek word for salt. The saliferous underground strata quickly closed up after the abandonment of an mine alit (passage- way), excellently preserving materials that usually are not durable (Beckel 1983). Due to modern salt mining, documented since 1311 AD, many prehistoric shafts were destroyed. However, numer- ous others were rediscovered, thus uncovering many objects made of leather and wood, as well as fabric remnants. 14C AMS measurements were performed on spruce logs from 3 different rediscovered mines to determine when salt mining began in Hallstatt.

SAMPLE PREPARATION

We describe below the sample preparation procedures at the 3 AMS laboratories involved in the Ice- man measurements.

Vienna, Austria First, the samples were cleaned mechanically by removing adherent particles with a scalpel and cleansing with bidistilled water in an ultrasonic bath. Next, the AAA (acid-alkali-acid) method was applied to the Iceman and Hallstatt samples (see e.g. Bonani et al. 1994; Wild et al. 1998). All the samples except 2 were successively treated with 1 M HCI, 0.1 M NaOH and 1 M HCI. Each pretreatment step was performed for 1 h at 60 °C; between each step, and at the end, the samples were washed to neutral pH with bidistilled water. Our standard AAA method described above was slightly modified for 2 Iceman samples to avoid dissolution of the sample material. VERA-0056, a hair sample from an ibex (wild goat; see Table 1), was kept at room temperature during the alkaline step, and for VERA-0054 (leather, see Table 1), 0.01 M NaOH instead of 0.1 M NaOH was used and the entire AAA method was performed at room temperature. The blanks (specular graphite containing no 14C) and the IAEA standard materials C-3 cellulose and C-5 wood were also subjected to our standard procedure. The IAEA standard material C-6 sucrose was not pretreated. After chemical pretreatment, the samples, standards and blanks (usually about 10 mg), together with some Ag wire, were put into a quartz vial containing 1 g Cu(II)O rodlets. The evacuated and flame- torch sealed vials were heated in an oven at 900 °C for 4 h. Following the method of Vogel et al. (1984) the resulting CO2 was transferred to a graphitization system where the CO2 was catalytically reduced (< with high-purity H2 to elemental carbon on an iron powder catalyst 200 mesh) according to: (Fe, 580 °C) CO2+2H2-> C+2H2O (1)

The carbon-catalyst mixture of each sample was split up into 3 or 4 portions containing about 1 mg of carbon. They were pressed into Al target holders (drill holes 1.1 mm in diameter) with a recess of 0.5 mm. Two or 3 of them (i.e., duplicates or triplicates of the same sample) were used for the AMS measurements; the remaining targets were stored in an archive.

Uppsala, Sweden Samples were cleaned mechanically first, removing visible particles, etc. Next, an AA (acid-alkali) method was applied: 1)1% HCl for 6-8 h kept below the boiling point, 2) washing in distilled water, Dating Equipmentfrom the Iceman 185

3)1% NaOH for 6-8 h kept below the boiling point, and 4) washing in distilled water. The remaining insoluble fraction was combusted with Cu(II)O at 800 °C for 10 min and graphitized using Fe and H2 at 750 °C. Each of the 2 grass samples (Ua-2373 and Ua-2374) was pretreated uniformly, but 2 independent combustions and graphitizations for each sample were done (G Possnert, personal communication 1992).

Gif-sur-Yvette, France

All samples were divided into 2 or 3 subsamples, to which 2 different pretreatment procedures were applied: l) an AAA method similar to the one applied in Vienna, and 2) a cellulose extraction method for the wood samples replacing the last acid step in step 1) by bleaching in NaClO2 at 80 °C. Each subsample was also combusted and graphitized separately using Fe as a catalyst. The carbon-catalyst mixture yielded 2 or 3 targets with 1 mm diameter (M Arnold, personal communication 1998). 14C AMS measurements

In this paper, only the measurements performed at the VERA 3 MV Pelletron® accelerator (Kut- schera et al. 1997; Priller et al. 1997) in Vienna are described in detail. Datings of the 2 grass samples Ua-2373 and Ua-2374 (previously published in Prinoth-Fornwagner and Niklaus 1994, but given here in much more detail in Table 1) were performed at the 7 MV EN tandem accelerator in Uppsala, Sweden (Possnert 1984). Another grass sample, l tree leaf sample and 6 wooden samples from the vicinity of the Iceman (see the GifA laboratory numbers in Table 1) were measured at the 3 MV Tandetron® accelerator in Gif-sur-Yvette (Arnold et al. 1987).

The datings at VERA comprised all the other Iceman samples specified in Table 1 as well as the wooden samples from Hallstatt shown in Table 2a. Samples were loaded into the 40-position target wheel of the Multi-Cathode SNICS sputter source (each position contains one target). 13C/12C and 14C/12C ratios were obtained by using the 3+ charge state at 2.7 MV terminal voltage. For details of the machine performance see Priller et al. (1997) and Rom et al. (1998).

The sample loading contained 9 samples from the Iceman (6 triplicates and 3 duplicates), 3 different IAEA standard materials (C-6 sucrose triplicates, C-5 wood duplicates, and C-3 cellulose triplicates) warranting a high level of quality control, 3 machine blank targets, i.e. "dead" graphite con- 14C, taining no and a chemistry blank duplicate, i.e. a machine blank pretreated and processed in the same way as the samples. The Hallstatt sample loading contained 10 samples (8 triplicates and 2 duplicates), C-6 sucrose and C-3 cellulose triplicates, 2 machine blanks and a chemistry blank qua- druplicate. Each target was measured for 200 sec, 8 times. Two of the samples (VERA-0040 and VERA-0041) were measured in 2 consecutive target wheels. The upper limit for the error on counting statistics was about 0.5% for every target except blanks.

As a quality control of the chemistry processing and the machine setup in all the regular 14C dating measurements performed so far at VERA, a C-6 Sucrose standard has been compared to a C-3 Cel- lulose standard measured in the same target wheel (see Figure 4). The corresponding chemistry 14C blanks provide the absolute dating limit Tmax for each laboratory. This limit is connected to the scatter of the blank L\pMCblank given in percent modern carbon (pMC, see Stuiver and Polach [1977]) via

2 L\PMCblank T = -8033 In (2) max 100

(Compare with Mook and Streurmann [1983] or Donahue [1990a].) 14C Table 1 Summary of dating results for samples taken from the Iceman's equipment Archaeo- Dry Chemical Calibrated logical Detailed weight pretreat- 14C age x2-test ranges age find nr Lab nr Specification Species localization (g) ment' (yr BP) BC)` B-91/166 GifA-91413 Leaves, parallel Grasses Grasses from the 0.09 AAA 4550 ± 60 sample to no. (Poaceae) cape B-91/16a 3383030 0.93

B-91/32 GifA-93033 Stiffening of the Hazel Inner parts of the 0.04 AAA 4690 ± 70 GifA-93034 quiver (Corylus avellana) notch Cellulose 4620 ± 60 -27e GifA-94367 Cellulose 4460 80 -21e average t 4605 ± 40 -26 4.82 / 5.99 3220-3100 0.15 B-91/33 GifA-93035 Wood from the Hazel Hardwood from 0.05 AAA 4430 60 GifA-93036 t pannier (Corylus avellana) the broken end of Cellulose 4680 ± 60 -30 Q GifA-94368 the hazel stem Cellulose 4500 ± 80 -19 e average 4540 t 70 -26 9.05/5.99 B-91/34 GifA-93038 Wood from the Hazel Wood from the 0.06 AAA 4710 t 70 GifA-93039 pannier (Corylus avellana) broken end of the Cellulose 4670 t 60 -32 e GifA-94369 hazel stem Cellulose 4480 t 90 -27 e average 4645 ±40 -30 4.40/5.99 3520-3350 0.98 B-91/35 GifA-93040 Wood from the Yew Splinter from the 0.13 AAA 4540 t 70 e GifA-93041 bow (Tarus baccara) broken end Cellulose 4700 ± 70 -27 e GifA-94370 Cellulose 4530 ± 80 23e average 4595 ±40 -26 3.53/5.99 33903310 0.38 3240-3100 0.26 B-91/36 GifA-93043 Wood from the Yew Wood from the 0.04 AAA 4440 ± 60 GifA-93044 axe-shaft (Taxus baccata) inner sides of the Cellulose 4500 ± 70 -25 e GifA-94371 splitted branch Cellulose 4450 t 70 27e average 4460 ± 40 -28 0.46/5.99 2982920 0.04 B-91/37 GifA-93045 Wood from the Haze] Splinter from the 0.21 AAA 4430 t 70 GifA-93046 pannier (Corylus aveUana) broken end Cellulose 4540 ± 50 -28 e GifA-94372 Cellulose 4420 ± 70 -31 e average 4480 ± 40 -30 2.68/5.99 B-91/38 GifA-93047 Leaves from the Norway Maple Tree Found nearby the 0.08 AAA 4540 ± 70 ember vessel (Acer platanoides) quiver aAA = acid-alkali AAA acid-alkali-acid Cellulose = = acid-alkali + NaClOa pretreatment. bx2-values, corresponding 95% confidence limits for 3 subsamPles of the sample, which were independently chemically Pretreated and graphitized. Calibration by OxCal v2.18 with default system options and the INTCAL98 calibration curve; 95.4% confidence ranges (2 6 . dRelative probability of finding the true age in the respective time range. The absolute probability is obtained by multiplying by 0.954 (2 6 range). eError estimated to 3%. 14C Table 1 Summary of dating results for samples taken from the Iceman's equipment Continued Archaeo- Dry Chemical Calibrated logical Detailed weight Pretreat- 14C age x2-test ranges age find nr Lab nr Specification Species localization (g) menta (Yr BP) BC)C B-91/3 Ua-2373 Leaves Grasses Grasses from the 0.13 AA 4620 t 75 (Poaceae) filling of his left 4605 ± 70 -25e shoe 4612 t 51 0.02/3.84' 3650-3100 B-91/16a Ua-2374 Leaves Grasses Grasses from the 0.07 AA 4250 t 70 (Poaceae) cape 4450 t 75 -25e 4343 ± 100 3.80/3.84' > 3350-2650 B-91/35 VERA-0050 Wood from the Yew Recovered during AAA 4500 ± 30 t 1.2 3.84 "> bow, parallel (Taxes baccata) sample washing sample to no. B-91/35

91/96 VERA-0051 Wood Hazel Recovered during 30 1.2 5.99 11> (C. avellana) sample washing 91/139 VERA-0053 Charcoal from Conifers Recovered during 0.27 AAA 4690 ± 40 -23.0 ± 1.9 0.08 / 3.8411 3633580 0.15 the ember vessel sample washing 3543360 0.85 from the meltwa- ter channel 91/139 VERA-0049 Leaves from the Norway Maple Recovered during 0.06 AAA 4510 ± 40 -27.2 ± 1.8 0.11 / 3.8410 3360-3090 0.98 ember vessel, Tree (Acer sample washing 3060-3040 0.02 parallel sample to platanoides) from the meltwa- no. B-91/38 ter channel 92/181 VERA-0054 Leather Recovered during 0.07 AAA 4480 ± 40 -22.1 ± 1.4 0.57 / 59911) 3353020 1.00 sample washing 92/275 VERA-0048 Wood, binding Green Alder Found in the west- 0.04 AAA 2500 t 40 t 1.8 59911) material (Alnus viridis) em part of the southern rock-rib 92/283 VERA-0056 Hair Sediment from the 0.49 AAA 4510 t 30 t 1.7 59911) (Capra ibex) gully 92/283 VERA-0055 Mosses Sediment from the 0.08 AAA 4700 ± 40 t 1.6 3.84" sexangulare gully 3543370 0.78 92/292 VERA-0052 Wood Pine sp.) Found on the 0.02 AAA 5820 ± 40 t 1.5 59911) southern rock-rib aAA 1) = acid-alkali AAA = acid-alkali-acid, Cellulose = acid-alkali + NaClO2 Pretreatment. bx2-values and corresponding 95% confidence limits for two independent graPhitiza- tions of the same chemically uniformly Pretreated sample; u) several targets of a uniformly Pretreated and graPhitized sample. Calibration by OxCal v2.18 with default system options and the INTCAL calibration curve. The 95.4% confidence ranges (2)6 are given. Given 14C ages were rounded according to (Stuiver and Polach 1977) before calibrating. dRelative probability of finding the true age in the respective time range. The absolute probability is obtained by multiplying with 0.954 (2-6 range). e 813C estimated> not measured. J00 Table 2a. Summary of 14C results for samples taken from logs in the salt mines of Hallstatta Archaeological Location identification nr Dendro- (mining Weight age x2-test age ranges of age Lab nr (inventory nr) identification Site group) Species (g) (Yr BP) (95%)b (Yr BC)C ranged

VERA-0040 Hallstatt 1997/07 Log #1, outermost GrGner mine Northern Spruce 0.071 3275 ± 35 0.7 (77391) rings (58-66) 3220± 60 -20.4 ±0.8 3260± 35 -19.7 ±0.5 0.63/3.84' VERA-0041 Hallstatt 1997/08 Log #1, innermost Griiner mine Northern Spruce 0.183 3080 ± 35 (77391) rings (1-10) 2970± 60 -28.4 ±0.7 3060 ± 45 -26.6±0.5 2.58/3.84' > 1140-1130 0.01

VERA-0042 Hallstatt 1997/09 Log #2, outermost Gruner mine Northern Spruce 0.102 3040 ± 40 0.9 3.84 ") (73232D) rings (409) 1141130 0.02

VERA-0039 Hallstatt 1997/06 Log #3, outermost Gruner mine Northern Spruce 0.112 3020 ± 45 1.5 3.84 "> (88887) rings (68-76)

VERA-0043 Hallstatt 1997/10 Log #4, outermost Log house 40 0.7 5.99 ") (68795C) rings (71-80)

VERA-0035 Hallstatt 1997/02 Log #5, innermost Appold mine Northern Spruce 0.371 3035 ±40 1.2 5.99 II> (4842G) rings (1-10) 1140-1130 0.03

VERA-0034 Hallstatt 1997/01 Log #5, outermost Appold mine Northern Spruce 0.140 3040 ± 45 1.3 5.99 « (4842G) rings (41-50)

VERA-0034 (after "wiggle matching" log #5 with VERA-0035 and VERA-0034) 1371170 1.00 VERA-0038 Hallstatt 1997/05 Log #6, innermost Tusch mine Eastern Spruce 0.256 2945 ±35 1.1 5.99 °> (F'Nr.90) rings (1-8)

VERA-0037 Hallstatt 1997/04 Log #6, intermedi- Tusch mine Eastern Spruce 0.156 2940 ± 40 1.4 5.99 «> (FNr.90) ate rings (36-45)

VERA-0036 Hallstatt 1997/03 Log #6, outermost Tusch mine Eastern Spruce 0.692 2920 ± 50 ±2.3 5.99 "> (F'Nc.90) rings (70-80)

VERA-0036 (after "wiggle matching" log #6 with VERA-0038 and VERA-0036) 1190-1010 1.00 aAll samples received AAA pretreatment (acid-alkali-acid). bx2_values and corresponding 95% confidence limits for I) two independent measurements of targets of the same chemically uniformly Pretreated sample mounted into different target wheels; II) several targets of a uniformly Pretreated and graPhitized sample. Calibration by OxCal v2.18 with default system options and the INTCAL98 calibration curve. The 95.4% confidence ranges (26) are given. dRelative probability of finding the true age in the respective time range. The absolute probability is obtained by multiplying with 0.954 (2-6 range). Dating Equipmentfrom the Iceman 189

DATA EVALUATION AND CALIBRATION

Although AMS measurements were performed at 3 different AMS laboratories, only the data evaluation at VERA shall be discussed in detail (see also Kutschera et al. (1997); Priller et al. (1997); Rom et al. (1998)).

First, for each target j, the mean values of the 13C/12C (R131) and 14C/12C (R143) ratios, respectively, and the corresponding standard deviations of the mean (S13j, S14j) were computed: 1 1 R13=J R13;k R14J = k 3b) m , yyl R14(3a,, =1 Jk. =1 k

nt . m. 1 2 2 (R131k - R13j) (R14j k - R14j) k=1 k=1 S13j = S14j = (4a, 4b) N mj(mj -1) N mj(mj-1)

13C/12C kth jth where R133,k denotes the ratio of the measurement on the target, R14j k the correspond- 14C/12C ing ratio, m1 is the number of measurements performed at target j, 5131 stands for the standard deviation of R131, S14j for the standard deviation of R14j.

The error for each target originating in counting statistics only can be written as

m, 3 2 (Sstatj k) k=1 R14j, k Sstatj = with Sstatj = (4c) mj k 12,,3+ Cj k

14C/12C kth where Sstatj k denotes the counting statistical error of the ratio of the measurement on the jth target, 12C3 ,k is the corresponding number of 1203+ ions. In the next step, for all targets containing the same sample material, the weighted mean R of the 13C/ 12C 14C/12C (R13) and the (R14) ratios, respectively, and the corresponding internal and the external errors (Sint and Sext) are computed. When calculating these weighted means, for the 13C/12C ratios 14C/12C Si in (5) is replaced by S 13j, in the case of ratios Si is replaced by the larger of the two errors S 14 and Sstatj, n is the number of targets containing the same target material.

n '352R 1 R - (5) n

n 1 (R _R)2 2(R _ R)2 s. S1 ()2J j 3 1 j=1 j =1 S int Sext S l n t (ba, 6b) = n ; = n (n-1)L 1 (S . )2 N = 1 j 190 WRom et al.

For the 613C corrections we use the following definitions

_ R13samle - R13PDB R13d - R13PDB &3Csampie - 1000 %o , &3CStd = 1000 %o (7a, 7b) R 13 PDB R 13 PDB where the subscript PDB denotes the PDB standard and std the standard material (usually C-3 cellulose) to which our measurements are referenced. The b13Cstd is the recommended IAEA value of this particular standard material.

Eliminating R13PDB results in

13 13C [R13samie( b Csrd _ 1 + _ 1 1000 %. ( 8 ) sample R13std 1000

The corresponding error is obtained by using Gauss's equation for propagation of errors considering correlations.

Finally, the 14C3+/1203+ ratios of the samples, standards and blanks are normalized to b13C = -25% (Stuiver and Polach 1977) by the quadratic form of the fractionation correction (see Stuiver and Robinson 1974)

R14_25 x = R14 x (9)

where the subscript x holds for sample, std or blank. The following background correction was used:

(R14sample, -25 -R14blank, -25) pMCsample = pMCrecommendedstd (10) (R14std -25 - R14blank _25) pMCsrd ommended is the pMC value recommended by the IAEA for the particular standard material. The corresponding error again is obtained by using Gauss's equation for propagation of errors considering correlations.

This correction is valid for the following assumptions: a) the carbon masses of the sample, the standard and the blank material are all the same, and the contamination is equal for all those materials (in mass and in isotopic composition), or the carbon masses for all these materials may be different, but the contamination is strictly proportional to the mass; b) the portion of extraneous modern carbon, i.e. carbon with an 14C/12C ratio of about 1.210-12 or 100 pMC, introduced as contamination is small compared to the sample, to the standard and to the blank size. This contaminating carbon is known to be in the range of a few µg (Vogel et al. 1987; Donahue et al. 1990b), compared to a few mg target size. From the data shown in Figure 4c it can be deduced that for samples at the 5-mg level, typically processed at VERA, the corresponding contamination amounts to about 15 µg modern carbon. Dating Equipment from the Iceman 191

It is interesting to note that equation (6b) can be rearranged as

n 1 R) 2 (Sext)2 a(Ri - = 2(n - I) (11) =1(S) (S int)

Assuming that the RJ are independent, normally distributed random variables with mean value R and standard deviations SJ, the sum on the left side is x2-distributed with (n-1) degrees of freedom (see e.g. Martin 1971; Ward and Wilson 1978). Therefore, one can use this sum in a Z2-test for checking the hypothesis Ho that the data R stem from the same normally distributed parent population with mean value R vs. the alternative hypothesis H1. The R all can have different standard deviations S reflecting different levels of precision in measuring the data. If the limit for a given confidence level (usually 95%) is exceeded by the sum in (11), H1 is preferred to Ho. One should keep in mind that 14C ages are in principle not normally distributed (like the pMC values), but if the ages to be averaged do not differ too much the Z2-test can be applied directly to the BP values (see Ward and Wilson 1978).

Tables 1 and 2a show the results for all measurements of the Iceman and the Hallstatt samples. Table 2b gives a comprehensive summary of already published dates from the beginning of salt mining in Hallstatt. Conversion of all the 14C ages to calendar ages given in this paper and also the "wiggle matching" of the spruce log samples from Hallstatt have been performed via the calibration program OxCal v2.18 (Ramsey 1995a, 1995b) using the default system options (cubic smoothing of the calibration curve, rounding off of the calibrated age ranges, etc.) and the INTCAL98 14C calibration curve (Stuiver et al. 1998).14C ages were rounded off according to Stuiver and Polach (1977) before calibrating. The so-called "wiggle matching" (referring to "wiggles" in the calibration curve, which originate in the varying 14C content of the atmosphere) is a useful procedure in combining dates from several pieces of wood stemming from the same log. Since the difference in calendar age between this pieces is known via the tree-ring sequence, this additional information can be used to obtain more accurate dates (Ramsey 1995b).

Table 2b. Summary of already published 14C results for samples taken from the oldest parts of the salt mines of Hallstatt Location Calibrated Fraction (mining Specific 14C age ranges age Lab nr Site group) BP) BC)a GrN-19975 Gruner mine Northern - 13 (1993-94) GrN-6047 Gruner mine Northern - 35 (1974) VRI-345 Gruner mine Northern handle 90 (1974) 1220-800 0.99 GrN-19842 Graner mine Northern - 35 (1993-94) VRI-267 Flechner mine Northern 100 (1973) GrN-19840 Tusch mine Eastern - 50 (1993-94) GrN-19973 Tusch mine Eastern - 16 (1993-94) VRI-558 Kaiser-Josef-alit 100 (1973) (Kilb mined) VRI-258 Kaiser-Josef- Eastern frag- 90 (1973) Querschlag I ment of a tool 1220-800 aCalibration by OxCal v2.18 with default system options. The 95.4% confidence ranges (2 6) are given. Given 14C ages were rounded according to (Stuiver and Polach 1977) before calibrating. bRelative probability of finding the true age in the respective time range. The absolute probability is obtained by multiplying with 0.954 (2-a range). Marked as unexpectedly high in (Felber 1973) and discarded as a statistical outlier in (Barth 1993-1994). d5ee Barth (1993-1994). 192 W Rom et al.

14C All except 1 of the averaged ages shown in Table 1 are weighted averages. The corresponding errors for the Gif-sur-Yvette samples are internal errors, only if the 95% confidence limit of the above-mentioned Z2-test is exceeded the unweighted mean and the corresponding standard devia- 14C tion is given (see archaeological find number B-91/33 in Table 1). The averaged ages of the samples measured at Vienna and also of those measured at Uppsala were determined as described above using the larger of the internal and external errors.

AMS lab and sample material Uppsala: r Leaves, Grasses (Poaceae) B-91/3 - a) B-91 /16a r Leaves, Grasses (Poaceae) Gif-sur-Yvette: e B-91/16b r Leaves, Grasses (Poaceae) B-91 /32 r Stiffening of the quiver, Hazel (Corylus avellana) E B-91/33 r r Wood from the pannier, Hazel (Gory/us avellana) B-91/34 r Wood from the pannier, Hazel (Gory/us avellana) B-91/35 r Wood from the bow, Yew (Taxus baccata) B-91/36 r Wood from the axe-shaft, Yew (Taxus baccata) B-91/37 r r Wood from the pannier, Hazel (Gory/us avellana) B-91138 r Leaves from the ember vessel, Norway Maple Tree (Acer platanoides) Vienna: B-91/35 r r Wood from the bow, Yew (Taxus baccata) 91/96 r Wood, Hazel (Gory/us avellana) 91/139 r Charcoal from the ember vessel, Conifers 91/139 r Leaves from the ember vessel, Norway Maple Tree (Acer platanoides) 92/181 r Leather 92/275 a r Wood, binding material, Green Alder (Alnus viridis) 92/283 r Hairs, Ibes (Capra ibex) 92/283 r Mosses (Polytrichum sexangulare) 92/292 r r Wood, Pine (Pinus sp.) a) 1 1 1 1 1 1 1 1 1 parallel 1 1._1 1 1 1' 1 1 111 1 1' 1' samples 5000 4000 3000 2000 1000 0 Calendar age (years BC)

Figure 1 The "Iceman" samples. Horizontal bars indicate 95.4% confidence ranges (26), the dashed vertical lines show the 2-6 range obtained from tissue and from bone samples measured at the AMS laboratories of Zurich, Swit- zerland and of Oxford, England (see text).

THE ICEMAN

Table 1 and Figure 1 show the results of the Iceman samples that originate from 2 excavations in the vicinity of "Otzi" in 1991 and 1992 (Lippert 1992; Bagolini et al. 1995). The grass samples were taken at the Institute for Forensic Medicine at the University of Innsbruck 1 week after the salvage of the Iceman; all the other samples were taken at the Romisch-Germanisches Zentralmuseum in Mainz in 1991 and 1997. The samples consist mainly of wooden material, but also mosses and leather were dated. The bulk of these samples confirms the time period of 3360-3100 BC previously determined from bone and tissue specimens from the Iceman (Bonani et al. 1992, 1994; Hedges et al. 1992; Prinoth-Fornwagner and Niklaus 1994)1. However, for 2 wooden samples, clearly deviating dates were found: 790-410 BC (VERA-0048) and 4790-4550 BC (VERA-0052). These samples provide evidence that the site of the Iceman was used as a mountain pass 1500 yr earlier, i.e. at the transition of the Mesolithic to the Neolithic period, confirming an archaeological presumption expressed earlier (Bagolini et al. 1995), and 2000 yr later, i.e. in the Hallstatt period (Figure 1, Table 1). Especially interesting is the younger sample, showing clear working traces (K Oeggl, personal communication 1998). It is the first artifact from the Iron age found in this Alpine region (Kutschera et al. 1998). Dating Equipment from the Iceman 193

Dendro-identification Lab number (?) log #1, outermost rings (58-66) - VERA-0040 log #1, innermost rings (1-10) Gruner VERA-0041 log #2, outermost rings (40-49) mine VERA-0042 log #3, outermost rings (68-76) - VERA-0039

log #4, outermost rings (71-80) log-cabin] - VERA-0043

log #5, innermost rings (1-10) - VERA-0035 log #5, outermost rings (41-50) Appold mine - VERA-0034 log #5, "wiggle matched"

log #6, innermost rings (1-8) VERA-0038 log #6, intermediate rings (36-45) Tusch VERA-0037 log #6, outermost rings (70-80) mine VERA-0036 log #6, "wiggle matched"

1700 1600 1500 1400 1300 1200 1100 1000 900 Calendar age (years BC) Figure 2 Hallstatt samples from the earliest salt mining period. The measured samples originate from different salt mines (Gruner, Appold, Tusch) at Hallstatt, and from a log cabin found nearby. 2- 6 ranges of single calibrated 14C dates from tree-ring samples averaging about 10 consecutive years are shown. The somewhat reduced time ranges (1-c and 2-6 ranges given) were obtained by "wiggle matching" of these dates for log #5 and #6, respectively, using the calibration program OxCal v2.18 (Ramsey 1995b). They represent the calibrated ages for the respective outermost tree rings being closest to the event to be dated. For log #1 no "wiggle matching" was performed (see text).

The first dates on botanical remains were published in Egg and Spindler (1992). However, information on these findings (e.g. botanical specification) is incomplete. Also, no reference to the calibration curve used and the calibration procedure is given. '4C ages are not given either. This makes it difficult to include these dates in a general comparison. Nevertheless, for completeness, we list these results here: Uppsala/S 3053-2931 BC (16) corresponds to Ua-2374 (G Possnert, personal communication 1992) (see Table 1), Paris/F 3365-3106 BC (16) is likely to be GifA-91413 (Table 1), Cam- bridge/Mass. USA 3362-3136 BC (16), 3492-3049 BC (2 6) presumably is GX-18504-AMS (see Prinoth-Fornwagner and Niklaus 1994 and below).

Sample GifA-91402 published in Prinoth-Fornwagner and Niklaus (1994) should be discarded, since according to M Arnold (personal communication 1995) the sample burnt was very small and the contamination was certainly larger than normal. A re-calibration of 2 further grass samples, ETH-8345.3 (4535 ± 60 BP, b13C = (-25.4 ± O.9)%, 1.5 mg sample size, AAA pretreatment)

IA combined calibration of the weighted means for tissue (4523 ± 27 BP) and bone samples (4576 ± 27 BP) determined at the AMS laboratories of Zurich and Oxford was performed using a time shift of 20 yr for the mean carbon turnover time in bone collagen (see Bonani et al. 1992; Hedges et al. 1992). Calibration yields time ranges of 3350-3320 BC (0.72), 3190- 3170 BC (0.12), 3140-3120 BC (0.17) for the 68.2% confidence interval, and 3360-3310 BC (0.63), 3200-3100 BC (0.37) for the 95.4% confidence interval; the numbers in brackets denote the relative probability of finding the true age in the respective time range. 194 W Rom et al.

(Bonani et al. 1994) and GX-18504-AMS (4555 ± 48 BP, b13C = -25.5%o) (Prinoth-Fornwagner and Niklaus 1994), yields 3500-3460 BC (0.03), 3380-3020 BC (0.97) and 3500-3460 BC (0.06), 3380-3090 BC (0.94) for the 95.4% confidence intervals; the numbers in brackets following calibrated age ranges denote the relative probability of finding the true age in the respective time range.

Lab number - r VERA-0041 - GrN-19975 - r VERA-0042 - Gruner mine VERA-0039 - GrN-6047 VRI-345 GrN 19842

log-cabin VERA-0043

- VERA-0034 Appold mine (after "wiggle matching")

w; i':.:'::>i:c :s: v+i;.:: ::::'. :ti. i; vy+i.:iii';0; :: .... i?'i 4?:;:..:Yi+i; Li:'i; is::.:.: .....:....:...... :...... +.., ...... M.iw Flechner mine VRI-267

- GrN-19840 Tusch mine GrN-19973 r VERA-0036 (after "wiggle matching")

Kaiser-Jose f ::::;;;:::::;:::::::...:.::.:::::.:::...... :::::.:.:: , ... vRI-258 Querschlag 1

1400 1300 1200 1100 1000 900 800 700 600 Calendar age (years BC)

Figure 3 Comprehensive survey of 14C dates of wooden objects from the beginning of salt mining in Hallstatt. Both 1-6 and 2-6 ranges for single calibrated 14C dates are shown. VRI-558 and VERA-0040 have been omitted from this figure as outliers (see footnotes in Table 2b and text, respectively). VRI = Vienna Radium Institute, GrN = Centrum voor Isotopen Onderzoek, Groningen.

THE SALT MINES OF HALLSTATT

The 14C dating of objects presumably originating from the beginning of salt mining in Hallstatt, comprised spruce logs from 3 different prehistoric salt mines and a nearby log cabin on the surface. Two of these mines, the "Appold" and the "Gruner", were rediscovered in 1879 and 1910, respectively (Barth 1993; Barth and Neubauer 1993). They are located in the Northern mining group, which is considered to be the oldest part of the salt mining field of Hallstatt. The 3rd mine, the "Tusch", was first discovered in 1748 and then rediscovered in 1991(Barth 1993- 1994). Two further samples shown in Figure 3 from the "Flechner" mine and the "Kaiser-Josef- Querschlag I" also belong to the oldest objects found in the salt mines of Hallstatt. The wane (Wald- kante) of all logs shown in Figure 2 and Table 2a was still preserved. This enables a direct dating of the felling of the trees without any age offset. Usually, the timber was used only once (presumably) without any preceding storage (P Stadler, personal communication 1999). For all the other samples from Hallstatt (Table 3) no information concerning the trimming of the timbers or the presence of bark or sapwood has been published.

The 14C dates from the "Tusch" mine (Figures 2 and 3, Table 2a) including "wiggle matching" with OxCal v2.18 (Ramsey 1995b) set the beginning of the extraction of salt to the Late Bronze Age, i.e Dating Equipment from the Iceman 195

the Early Urnfield period. This agrees with data from the same mine published earlier in Barth 1993-94 (see Table 2b, Figures 2 and 3). However, from the 14C dates of findings in the Northern mining group (including "wiggle matching") one can infer that salt mining had already started in the Northern mining group in the 14th-13th century BC (see the "Gri ner" and "Appold" mines in Figures 2 and 3, and Tables 2a and 2b). This corresponds to the transition of the Middle to the Late Bronze Age, i.e. the transition from the Tumulus to the Urnfield period (Lippert 1985). Sample GrN- 19975 (see Figure 3) from the "Gruner" mine, which has been discarded as a statistical outlier in Barth 1993-1994, is now well confirmed by the dates obtained at VERA. Therefore, the beginning of salt mining in Hallstatt has to be set 1 to 2 centuries earlier than supposed so far (Barth 1993-94). The even higher age of 1 tree-ring set (see VERA-0040 in Figure 2) remains unexplained, but may reflect some non-removable contamination and therefore resists a reliable "wiggle matching".

156 154 IAEA C-6 Sucrose a) U 152 150 148 II 146 recommended IAEA value: (150.61 t 0.11)pMC 144 measured value: (150.42 t 1.09)pMC 0 :IAEA C-6 Sucrose b> 0 -5 -10 t -15 M -20 recommended IAEA value: (-10.80 ± 0.49)% -25 measured value: (-9.2 t 2.6)% Chemistry blanks c)

measured value: (0.30 t 0.05)pMC 1996 1997 1998 Year of measurement

Figure 4 Performance of standards and blanks in regular 14C-dating measurements. a) and b) over 12 months (April 1997-March 1998, symbols left of the vertical dotted line), the values measured for C-6 sucrose relative to C-3 cellulose agree well with the values recommended by the IAEA. The mean values shown are unweighted means, the corresponding errors are the standard deviations from that mean, c) Chemistry blank values. Taking the standard deviation of all the chemistry blank values of the last year (symbols left of the vertical dotted line) from the unweighted mean, our dating limit Tmax (see equation 1) now amounts to 55 ka. For clarity, an equidistant representation of data points is given in the figure, since the actual time spacing would tend to cluster the data.

CONCLUSION

Botanical remains from the vicinity of the Iceman, including his equipment, were dated at 3 different AMS laboratories. They show reasonable agreement with dates previously obtained on the tissue and bones of the Iceman himself. Two clearly deviating "outliers" indicate that the discovery site of Otzi had been used as a mountain pass 1500 yr before the lifetime of the Iceman, and also 2000 yr later. The latter date is particularly interesting, because it constitutes the first artifact from the Iron Age in the entire Otztal region. Spruce logs from the world's oldest salt mines at Hallstatt, Austria were subject to 14C dating with AMS. The results from VERA provide evidence that salt mining at Hallstatt started 1-2 centuries earlier than previously supposed, i.e. in the 14th to the 13th century BC. 196 W Rom et al.

A current project at VERA carried out in cooperation with the Institute for Botany in Innsbruck, Austria deals with more than 50 additional samples from the site of the Iceman discovery. Besides leather and droppings from animals, the samples include mainly botanical remains such as mosses, grasses, and saxifrages. 14C dating of high Alpine plants may give information about glaciation and possible deglaciation periods at the rock depression where the Iceman died. From these dates, indirect evidence for the climate of the last 5000 yr may be obtained.

ACKNOWLEDGMENTS This paper would not have been possible without the help of several colleagues who provided information and samples: M Arnold and N Tisnerat-Laborde from the Laboratoire des Sciences du Cli- mat et de l'Environnement in Gif-sur-Yvette, France, G Possnert from the Svedberg-Laboratoriet in Uppsala, Sweden, and S Bortenschlager and K Oeggl from the Institute for Botany in Innsbruck, Austria. We are especially indebted to F E Barth and P Stadler from the Museum of Natural History in Vienna, Austria.

REFERENCES Arnold M, Bard E, Maurice P, Duplessy C. 1987.14C dat- Gletscherleiche mit der Beschleuniger-Massenspek- ing with the Gif-sur-Yvette Tandetron accelerator: sta- trometrie-Methode (AMS). In: Hopfel F, Platzer W, tus report. Nuclear Instruments and Methods in Phys- Spindler K, editors. Der Mann im Eis 1. Report of the ics Research B29:120-3. 1992 International Symposium in Innsbruck, Publica- Bagolini B, Dal Ri L, Lippert A, Nothdurfter H. 1995. tions of the University of Innsbruck 187. 2nd rev ed. Der Mann im Eis: Die Fundbergung 1992 am Tisen- Eigenverlag der Universitat Innsbruck. p 108-16. joch, Gem. Schnals, Sudtirol. In: Spindler K, Rast- Bonani G, Ivy S, Hajdas I, Niklaus TR, Suter M. 1994. bichler-Zissernig E, Wilfing H, zur Nedden D, Noth- AMS 14C age determinations of tissue, bone and grass durfter H, editors. Der Mann im Eis: Neue Funde and samples from the Otztal Ice Man. Radiocarbon 36(2): Ergebnisse (The Man in the Ice, Volume 2). Wien and 247-50. W, K, W. 1992. New York: Springer-Verlag. p 3-52. Bortenschlager S, Kofler Oeggl Schoch Bahn BG, Everett K. 1993. Iceman in the cold light of Erste Ergebnisse der Auswertung der vegetabilischen day. Nature 362, 4 Mar:l l-2. Reste vom Hauslabjochfund. In: Hopfel F, Platzer W, Barfield L. 1994. The Iceman reviewed. Antiquity 68: Spindler K, editors. Der Mann im Eis 1. Report of the 10-26. 1992 International Symposium in Innsbruck, Publica- Barth FE, Felber H, Schauberger 0.1974. Radiokohlen- tions of the University of Innsbruck 187. 2nd rev ed. stoffdatierung der prahistorischen Baue in den Eigenverlag der Universitat Innsbruck. p 307-12. Salzbergwerken Hallstatt and Durrnberg-Hallein. Coghlan A. 1992.Otze: the man who came in from the Mitteilungen der Anthropologischen Gesellschaft in cold. New Scientist, 11 Jan: 17-8. Wien 105:45-52. Donahue DJ, Linick TW, Jull AJT. 1990a. Isotope-ratio Barth FE. 1993. Der urzeitliche Bergbau im Gri ner Werk and background corrections for accelerator mass spec- des Salzbergwerkes Hallstatt. In: Salinen Austria, ed- trometry radiocarbon measurements. Radiocarbon itors. Ausstellung Gruner Werk 1986. 2nd ed. Verlag 32(2):135-42. des Musealvereines Hallstatt. p 1-32. Donahue DJ, Jull AJT, Toolin U. 1990b. Radiocarbon Barth FE, Neubauer W. 1993. Salzbergwerk Hallstatt - measurements at the University of Arizona AMS fa- Appoldwerk - Grabung 1879/80. In: Salinen Austria, cility. Nuclear Instruments and Methods in Physics editors. Appoldwerk Grabung 1879/80.2nd ed. Verlag Research B52:224-8. des Musealvereines Hallstatt. p 3-53. Egg M. 1992. Zur Ausrustung des Toten vom Hauslab- Barth FE. 1993-94. Ein FUllort des 12. Jahrhunderts v. joch, Gem. Schnals (Sudtirol). In: Hopfel F, Platzer W, Chr. im Hallstatter Salzberg. Mitteilungen derAnthro- Spindler K, editors. Der Mann im Eis 1. Report of the pologischen Gesellschaft in Wien (MAGW) 123/124: 1992 International Symposium in Innsbruck, Publica- 27-38. tions of the University of Innsbruck 187. 2nd rev ed. Beckel L, Harl 0. 1983. Archaologie in Osterreich. Eigenverlag der Universitat Innsbruck. p 254-72. Salzburg and Vienna: Residenzverlag. (Reprint by Egg M, Spindler K. 1992. Die Gletschermumie vom Tosa Verlag, Vienna 1996). p 151-5. Ende der Steinzeit aus den Otztaler Alpen- Vorber- Bonani G, Ivy S, Niklaus TR, Suter M, Housley MA, icht. Jahrbuch des Romisch-Germanischen Zentral- Bronk CR, van Klinken GJ, Hedges REM. 1992. Al- museums Mainz 39(1):1-114. tersbestimmung von Milligrammproben der Otztaler Felber H. 1973. Altersbestimmungen nach der Radio- Dating Equipmentfrom the Iceman 197

kohlenstoffinethode am Institut fur Radiumforschung Prinoth-Fornwagner R, Niklaus TR. 1994. The man in the and Kernphysik IX. Anzeiger-Osterreichische Akade- ice: results from radiocarbon dating. Nuclear Instru- mie der Wissenschaften-Mathematisch-naturwissen- ments and Methods in Physics Research B92:282-90. schaftliche Klasse 110(1-13):57-65. Ramsey CB. 1995a. Radiocarbon calibration and analy- Hedges REM, Housley RA, Bronk CR, van Klinken GJ. sis of stratigraphy: the OxCal program. Radiocarbon 1992. Radiocarbon dates from the Oxford AMS sys- 37(2):425-30. tem: Archaeometry datelist 15. Archaeometry 34(2): Ramsey CB. 1995b. OxCal Program v2.18. URL: . Kutschera W, Collon P, Friedmann H, Golser R, Hille P, Roberts D, Garrett K, Harlin G. 1993. The Iceman. Na- Priller A, Rom W, Steier P, Tagesen S, Wallner A, tional Geographic, June:36-67. Wild E, Winkler G. 1997. VERA: A new AMS facility Rom W, Golser R, Kutschera W, Priller A, Steier P, Wild in Vienna. Nuclear Instruments and Methods in Phys- E. 1998. Systematic investigations of 14C measure- ics Research B 123:47-50. ments at the Vienna Environmental Research Acceler- Kutschera W, Golser R, Priller A, Rom W, Steier P, Wild ator (VERA). Radiocarbon 40(1):255-64. E, Arnold M, Tisnerat-Laborde N, Possnert G, Borten- Schauberger 0. 1960. Ein Rekonstruktionsversuch der schlager S, Oeggl K. Radiocarbon dating of equip- prahistorischen Grubenbaue im Hallstatter Salzberg. ment from the Iceman. In: Bortenschlager S, Oeggl K, Prahistorische Forschungen 5:1-15. editors. The Iceman and His Natural Environment. Spindler K. 1993. Der Mann im Eis. Munchen: C. Ber- The Man in the Ice 4. Vienna: Springer Verlag. Forth- telsmann Verlag GmbH. p 81-5, 93-6. coming. Stuiver M, Robinson SW. 1974. University of Washing- Lippert A. 1985. Reclams Archaologiefuhrer Osterreich ton Geosecs North Atlantic carbon-14 results. Earth and SUdtirol. Stuttgart: Philipp Reclam jun.. p 223-30. and Planetary Science Letters 23:87-90. Lippert A. 1992. Die erste archaologische Nachuntersu- Stuiver M, Polach HA. 1977. Discussion: reporting of chung am Tisenjoch. In: Hopfel F, Platzer W, Spindler 14C data. Radiocarbon 19(3):355-63. K, editors. Der Mann im Eis 1. Report of the 1992 In- Stuiver, M, Reimer, PJ, Bard, E, Beck, JW, Burr, GS, ternational Symposium in Innsbruck, Publications of Hughen, KA, Kromer, B, McCormac, G, van der Pli- the University ofInnsbruck 187. 2nd rev ed. Eigenver- cht J, Spurk, M. 1998. INTCAL98 radiocarbon age lag der Universitat Innsbruck. 245-53. calibration, 24,000-0 cal BP. Radiocarbon 40(3): Martin BR. 1971. Statistics for physicists. London and 1041-83. New York: Academic Press. p 55-61. Urban 0. 1989. Wegweiser in die Urgeschichte Osterre- Mook WG, Streurman HJ. 1983. Physical and chemical ichs. Vienna: Osterreichischer Bundesverlag. p 156- aspects of radiocarbon dating. In: Mook WG, Water- 63. bolk HAT, editors. PACT 8: Proceedings of the First Vogel JS, Southon JR, Nelson DE, Brown TA. 1984. Per- 14C International Symposium and Archaeology, Gro- formance of catalytically condensed carbon for use in ningen 1981. p 31-55. accelerator mass spectrometry. Nuclear Instruments Oeggl K. 1995. Neolithic plant remains discovered to- and Methods in Physics Research B5:289-93. gether with a mummified corpse ("Homo tyrolensis") Vogel JS, Nelson DE, Southon JR. 1987.14C background in the Tyrolean Alps. In: Kroll H, Pasternak R, editors. levels in an accelerator mass spectrometry system. Ra- Res archaeobotanicae - 9th Symposium IWGP. Kiel. p diocarbon 29:323-33. 229-38. Ward, GK, Wilson, SR. 1978. Procedures for comparing Possnert G. 1984. AMS with the Uppsala EN tandem ac- and combining radiocarbon age determinations: a cri- celerator. Nuclear Instruments and Methods in Phys- tique. Archaeometry 20(1):19-31. ics Research B5: 59-61. Wild E, Golser R, Hille P, Kutschera W, Puchegger S, Priller A, Golser R, Hille P, Kutschera W, Rom W, Steier Priller A, Rom W, Steier P, Vycudilik W. 1998. First P, Wallner A. Wild E. 1997. First performance tests of 14C results from archaeological and forensic studies at VERA. Nuclear Instruments and Methods in Physics the Vienna Environmental Research Accelerator. Ra- Research B 123:193-8. diocarbon 40(1):273-82.

RUDJER BOSKOVIC INSTITUTE RADIOCARBON MEASUREMENTS XIV

Nada Horvatincic Bogomil Obelic Ines Krajcar Bronic Dusan Srdocl Romana Calic Rudjer Boskovic Institute, PO Box 1016,10001 Zagreb, Croatia

INTRODUCTION

In this report we present dating of several large series of samples mostly collected in Croatia: l) Iron Age and medieval samples excavated in Zagreb, 2) speleothems from a karst site in central Croatia, 3) shallow sediments from the Plitvice lakes, 4) tufa from the Krka River, 5) sea sediments from the Adriatic Sea, and 6) tree rings from Hungary, Croatia and Slovenia. Sample preparation and proportional counter technique are essentially the same as reported earlier (Srdoc et al. 1971). Processing of data has been computerized (Obelic 1989). Age calculations follow the conventional protocol based on the Libby half-life (5570 ± 30 y), using AD 1950 as the reference year. Ages and standard deviations (16 error) of samples were adjusted for stable isotope fractionation according to the methods of Stuiver and Polach (1977). Calibrated ages (for archeological samples only) were calculated by using the program OxCal v.3.0 (Ramsey 1995, 1998) with 1c error (confidence level 68.2%). When several calendar age ranges were obtained, probability for each interval is given. Probabilities of <5% are omitted. Range intervals are rounded. 14C contents of geological samples are expressed in pMC, and of environmental samples in z 4C.

ARCHAEOLOGICAL SAMPLES Croatian Museum of Natural History Series

Wood fragments of a beam from remnants of a wooden house buried in the yard of the Croatian Museum of Natural History, Zagreb Upper Town, Croatia (45°50'N,16°0'E),150 m above sea level (asl). Collected and submitted in May 1992 by M Smalcelj, Department of Archaeology, University of Zagreb.

Comment: (MS) Conservation works in museum's yard. Expected age: 15th to 17th century. Z-2398. Croatian Museum of Natural History 1 200 ± 130 Wood fragment of a beam from western part of the house, 1.20 m depth (cal AD 1630-890, 58.9%; AD 1910-1950, 7.3%).

Z-2399. Croatian Museum of Natural History 2 220 ± 120 Wood fragment of a beam inside the house, 1.20 m depth (cal AD 1510-1590,13.4%; AD 1620- 1700,17.5%; AD 1720-1880, 24.4%).

Z-2400. Croatian Museum of Natural History 3 400 ± 90 Wood fragment of the floor, 1.50 m depth (cal AD 1430-1530, 42.1%; AD 1560-1640, 26.1%). Z-2401. Croatian Museum of Natural History 4 240 ± 90 Wood fragment of a beam outside the house, 1.80 m depth (cal AD 1510-1690, 43.1%; AD 1730- 1820, 21.5%).

1 Current address: Center for Radiological Research, College of Physicians and Surgeons, Columbia University, 630 West 168th Street, New York, NY 10032 USA

199 200 N Horvatincic et al.

Municipal Museum Series archaeological On the occasion of the 900th anniversary of Zagreb, the capital of Croatia, extensive Museum in the excavations were performed at the historical complex of the present-day Municipal Upper Town Upper Town of Zagreb, 161 m asl. The museum is located at the northeast corner of the built in the 17th and includes Popov Toranj (Priest's Watchtower), the former Convent of St Clare inner part and century (now the museum), and the Old Town granary. Samples were taken from the times surroundings of the former Convent of St Clare. The area has been inhabited since prehistoric in 14C (Iron Age), and was part of wooden fortifications during the Middle Ages. Minor differences are due to the ages between the previously published results (Obelic et al. 1995) and the present ones additional measurements. Samples were collected and submitted by M Smalcelj.

A) Parlatorium Boards below the present floor of the parlatorium of the convent. ± 100 Z-2448. Parlatorium 1 2085 Charcoal found beneath a Hallstatt bowl, quadrant P-4/C-2 (cal 360-310 BC, 7.8%; 210 BC-AD 20, 60.4%). Z-2449. Parlatorium 2 2040 ± 110 Charcoal from the same cultural layer as above, quadrant P-4/A-4 (cal 200 BC-AD 80, 68.2%). Comment: (MS) Estimated archaeological age: La Tene. Z-2450. Parlatorium 3 1775 ± 85 130- Same as above, but the age is not confirmed by archaeological finds; quadrant P-4/C-3 (cal AD 340, 68.2%). Z-2451. Parlatorium 4 1770 ± 85 Same as above, quadrant P-4/B-3 (cal AD 130-350, 68.2%).

B) Convent Hall the Remains of shallow sod houses overlain by dwellings with stone basements and stone hearths in eastern part of hall of the convent. ± 115 Z-2452. Convent Hall 1 2340 Fragment of dwelling post near fireplace; quadrant P-4/A-4 (cal 800-200 BC, 68.2%). Z-2453. Convent Hall 2 2100 ± 85 Fragment of dwelling post, layer 40 cm above fireplace, quadrant A-BI1-3 (cal 360-310 BC, 8.7%; 210 BC-AD 1, 59.5%). Z-2454. Convent Hall 3 2375 ± 105 Charcoal from fireplace, quadrant A-21 (cal 770-680 BC, 17.3%; 660-630 BC, 5%; 560-380 BC, 42.3%). Comment: (MS) Dated to an older Hallstatt phase. Z-2455. Convent Hall 4 1990 100 Charcoal from the remnants of a house, quadrant A-17 (cal 120 BC-AD 120, 67.2%). Z-2456. Convent Hall 5 2110 100 Charcoal from the remnants of a house, quadrant A-17 (cal 360-290 BC, 12.8%; 260-30 BC, 55.4%). Rudjer Boskovic Measurements XIV 201

Z-2457. Convent Hall 6 2590 t 130 Charcoal from the remnants of a house, quadrant A-16-17 (cal 900-750 BC, 30.9%; 720-520 BC, 37.3%).

Z-2458. Convent Hall 7 2505 t 80 Charcoal from the remnants of a house, quadrant B-12 (cal 800-740 BC, 12.5%; 730-520 BC, 55.7%).

Z-2459. Convent Hall 8 2290 # 105 Charcoal from the remnants of a house, overlying pit where sample Z-2457 was found, quadrant A- 14 (cal 520-190 BC, 68.2%).

C) Eastern Convent Wing

Samples below the floor of the eastern wing of the convent.

Z-2460. Eastern wing 1 1745 t 135 Burned oak from the remnants of a house, quadrant F-9.

Comment: (MS) Sample stratigraphically associated with La Tene, although not supported by 14C dating (cal AD 110-430, 68.2%).

Z-2461. Eastern wing 2 1630 t 105 Round logs from the remnants of a house, quadrant G-14.

Comment: (MS) Stratigraphically associated with La Tene, although not supported by 14C dating (cal AD 260-290; 6.8%; AD 320-550, 61.4%). Z-2463. Eastern wing 3 2255 t 110 Burnt trunk lying over the fireplace, quadrant A-14 (cal 400-160 BC, 65.9%).

D) Promenade Vrazovo Setaliste

Burnt timber from the remains of the rampart buried along the slope following the eastern wing of the convent to the present-day promenade Vrazovo setaliste. The burned layer consists of up to 50 cm of clay (average thickness 30 cm), overlying a timber foundation. The foundation seems to have been burned deliberately to fire the clay and provide a solid base ca. 6 m wide for an additional clay fill into which the wooden posts of a palisade were driven. The most probable period when the trees were felled was between the 4th and 3rd century BC (Obeli et al. 1995). Burnt Timber

Z-2482. Vrazovo etalite 1 2220 t 85 Quadrant K/L-37/38 (cal 390-190 BC, 68.2%). Z-2483. Vrazovo etalite 2 2350 t 85 Quadrant N-39; (cal 760-680 BC, 14.8%; 550-360 BC, 49.3%). Z-2484. Vrazovo etalite 3 2215 f 110 Quadrant K-38 (cal 400-160 BC, 65.7%).

Z-2485. Vrazovo etalite 4 2310 f 100 Quadrant K-38 (cal 550-200 BC, 65.5%). Z-2486. Vrazovo etalite 5 2250 t 75 Quadrant K/L-37/38 (cal 400-340 BC, 21.3%; 320-200 BC, 46.9%). 202 N Horvatincic et al.

Z-2487. Vrazovo etalite 6 2215 ± 90 Quadrant H-38 (cal 390-190 BC, 68.2%). Z-2488. Vrazovo eta1i3te 7 2300 105 Quadrant F-41/42 (cal 530-190 BC, 68.2%). Z-2490. Vrazovo eta1i3te 8 2400 140 Quadrant H-38 (cal 770-390 BC, 68.2%).

E) Chapter House Outer tree rings of boards, apparently used as roof shingle, thrown into a trench in the 15th century during the reconstruction of the Chapter house, which is located next to the parlatorium. 75 Z-2280. Chapter house 1 300 ± Fragment of board No. 12 (cal AD 1480-1660, 68.2%). Z-2285. Chapter house 2 260 ± 105 Fragment of board No. 26 (cal AD 1480-1690, 52.4%; AD 1740-1810,15.8%). Z-2286. Chapter house 3 325 ±80 Fragment of board No. 10 (cal AD 1480-1650, 68.2%).

F) Popov Toranj (Priest's Watchtower) Wood (Quercus sp.) from a beam, containing 50 tree rings, from ramparts located at the northwest corner of the convent complex. Z-2369. Popov toranj 1 765 ±50 Ten outer tree rings, top of the central post (cal AD 1220-1280, 68.2%). Z-2371. Popov toranj 2 540 ± 80 Ten innermost tree rings, same post as above (cal AD 1300-1360, 31.8%; AD 1380-1440, 36.4%).

G) Mediaeval Trenches

Samples from inner and outer trenches of the rampart. 115 Z-2405. Trench 1 480 ± Fragments of twigs buried in mud from profile of arch No. 2 (cal AD 1310-1360,15.3%; AD 1380- 1520, 47.2%, AD 1590-1630, 5.7%). Z-2406. Trench 2 230 ± 90 Wood buried in mud (cal AD 1510-1600,18.2%; cal AD 1620-1700, 20.3%; cal AD 1720-1820, 24.8%). Z-2407. Trench 3 540 80 Animal bone buried in mud (cal AD 1300-1360, 31.8%; AD 1380-1440, 36.4%). Z-2408. Trench 4 170 ± 85 Fragment of board from the arch No. 2, trench No. 2 (cal AD 1650-1700,13.6%, AD 1720-1880, 33.6%; AD 1836-1880,12.5%, AD 1910-1950, 8.5%). Z-2409. Trench 5 605 ± 80 Fragment of board (Quercus sp.) from the rampart's planking between pillars No. 4 and 5 (cal AD 1290-1400, 68.2%). Rudjer Boskovic Measurements XIV 203

Z-2410. Trench 6 440 ± 80 Fragment of wooden stick from Quadrant 29a, probably reconstruction of outer trench (cal AD 1400-1520, 57.6%; AD 1590-1630,10.6%).

Z-2411. Trench 7 580 ± 90 Fragment of damaged wooden door frame from the entrance to the trench next to the stairs, between pillars No. 14 and 15 (cal AD 1290-1420, 68.2%). Z-2412. Trench 8 795 ± 85 Fragment of wooden door from the entrance to the trench, next to the stairs, between pillars No. 14 and 15 (cal AD 1150-1290, 62.4%). Tree ring counting resulted in AD 1183-1289. (Durman, personal communication 1994).

Z-2413. Trench 9 430 ± 85 Fragment of a broom (Sorghum) from the entrance to the trench next to the stairs (cal AD 1400- 1530, 52.3%; cal AD 1580-1630,15.9%).

H) Convent Cloister

Wood samples from the cloister of the former convent. Comment: (Ms) Expected age: 16th century.

Z-2414. Cloister 1 260 ± 80 Wood fragment from quadrant A-10, Pit 2 (cal AD 1500-1680, 52.8%; AD 1740-1810,15.4%). Z-2415. Cloister 2 375 ± 85 Wood fragment buried in mud from Pit 3 (G-1) (cal AD 1440-1530,37.6%; AD 1560-1640, 30.6%). Z-2416. Cloister 3 280 ± 80 Wooden remains of a pillar buried in mud at W profile from Pit 2, corner of the cloister (cal AD 1480-1670, 65.0%). Z-2417. Cloister 4 Wood buried in mud below the colored eggshells, Pit 1(cal AD 1480-1640, 68.2%). Z-2425. Cloister 5 120 ± 80 Straw from Pit 3 (cal AD 1680-1750, 23.1%; AD 1800-1940, 45.1%).

GEOLOGICAL SAMPLES Hvar Island Series

Wood from seabed off Hvar Island, west of Korcula Island in the Adriatic Sea, southern Croatia. Seawater depth 74 m. Collected and submitted March 1994 by M Juracic, Geological Department, University of Zagreb.

Z-2507. Wood from seabed, depth 60-65 cm 18,700 ± 400 Z-2506. Wood from seabed, depth 120-125 cm 22,500 600 Adriatic Sea Series

Inorganic sediments from several boreholes from north Adriatic Sea, 74 km northwest from Pula (44°52'N, 13051'E), Istria, southwest Croatia. Investigation of natural gas reserves in the Adriatic Sea. Submitted April 1991 by E Prohic, Faculty of Natural Sciences, Zagreb. 204 N Horvatincic et al.

Table 1 Adriatic Sea series Sample nr Borehole Depth (m) Z-2333 J-1 725.0-725.5 0.8 Z-2334 725.5-726.0 7.0 ± 0.8 Z-2335 J-4 761.2-761.7 ±0.9 Z-2336 J-15/2 854.0-854.5 1.3 Z-2337 928.5-929.1 3.3 ± 1.2 Z-2358 930.5-931.1 8.4 ±0.6 Z-2338 931.5-932.0 9.0 ± 0.8 Z-2339 J-15/3 929.0-929.7 0.9 Z-2340 934.0-934.5 4.6 ± 0.5 Z-2341 936.0-936.5 7.5 ± 0.8 Z-2342 J-18/3 1019.5-11020.1 ±0.9 Z-2343 J-18/9 606.0-606.5 ±0.8 Z-2344 609.0-609.8 8.2 ±0.9 Z-2345 841.0-841.6 10.7± 1.1 Z-2346 1060.0-1060.8 3.9±0.8 Z-2347 Ivana-3 708.0-708.8 1.0 1-2348 710.0-710.8 4.5 ± 0.8 Z-2349 712.0-712.7 2.8 ± 0.8 Z-2350 715.0-715.6 <0.5 Z-2351 717.0-717.5 3.0±0.4 Z-2352 718.0-718.4 3.3 ± 0.6 Z-2353 722.0-722.5 3.4 ± 0.8 Z-2354 723.5-724.2 <0.5 Z-2355 728.5-728.9 14.1 ± 1.1 Z-2356 729.5-730.4 8.6 ± 1.0 Z-2357 733.5-734.3 15.4 ± 0.8

Tounj Series

Phreatic speleothems from Tounj Cave near Tounj (45°15'N, 15 °20'E), Dinarides, central Croatia. Samples collected from laminated speleothem at various distances from the bedrock. (Babic et al. 1996). Collected and submitted October 1992 by D Lackovic, Faculty of Natural Sciences and Mathematics, Univ. Zagreb.

Comment: (NH) 14C age does not correlate with stratigraphic position. Possible explanation: non- uniform contamination by detrital carbonate mud produced by weathering of Mesozoic carbonates. Krka River Series Tufa from Krka River between Knin (44°02'N, 16°l 1'E) and Skradin (43°49'N, 15°55'E), 56 km long, southern Croatia. The Krka River flows in the karst region of southern Dinarides and empties into the Adriatic Sea (Friganovic 1984). Tufa associated with moss, microscopic algae and cyano- bacteria forms the cascades in the Krka River, resulting in a string of waterfalls and lakes. Tufa samples were collected along streams and lakes and from old tufa deposits outside the present-day watercourse. Several old tufa samples were dated by the 230Th1234U method (Horvatincic et al. 1996). The sequence of tufa samples from Krka River National Park area follows the Krka River flow, from Krcic brook to Prukljan Lake. Samples collected and submitted in 1984 by D Srdoc and Rudjer Boskovic Measurements XIV 205

Table 2 Tounj series Distance from Sample nr base (cm) pMC Z-2429 0-0.3 11.5 ±0.5 Z-2478 0-1.8 6.0±0.4 Z-2434 1.5-1.8 8.9 ± 0.5 Z-2428 12.0-12.3 3.1±0.7 Z-2477 12.0-13.3 1.2±0.6 Z-2430 13.0-13.3 <0.5 Z-2514 65-66 4.4 ±0.5

B Obelic, in 1991 by D Srdoc, and 1996 by N Horvatincic and R. Calic. Ages of tufa samples from Krcic River were also published in Srdoc et al. (1984, 1987).

A) Krcic Series

Krcic brook empties into the Krka River near the Krka spring close to Knin.

Z-1322. Kric 1 0.8 pMC Porous tufa above Kric brook, right bank, in village of Krcic.

Z-1324. Kric 2 0.8 ± 0.7 pMC Hard tufa from Jejina Spilja cave, near Krcic brook.

Z-1326. Kric 3 0.9 ± 0.5 pMC Powdered tufa, ca. 12 m above Kric brook.

Z-1327. Kric 4 0.8 ± 0.5 pMC Hard tufa, top of the old tufa outcrop near Krcic brook.

Z-1328. Kric 5 7.6 ± 0.5 pMC Hard tufa, b13C below Z-1327, same tufa outcrop as above. 230Th/234U age 92 ±4 ka. = Z-1329. Kric 6 1.3 ± 0.5 pMC S13C = Hard, porous tufa, below Z-1328, same tufa outcrop as above. 230Th/234U age 201 ± 19 ka. Z-1330. Kric 7 6 ._.p8+ 0 7 MC Hard tufa in form of a petrified hollow trunk, same outcrop as above.

B) Topoljski Buk Waterfall Series

Topoljski Buk, a 22-m-high waterfall (also known as Kric waterfall), Kric brook, near its mouth in the Krka River; full of water in spring and autumn, and dry in winter and summer. Both recent and old tufa samples were collected at Topoljski Buk waterfall.

Z-1332. Topoljski Buk 1 57.5 ± 1.0 pMC Tufa from barrier, in a tunnel under barrier to Krka River spring.

Z-2360. Topoljski Buk 2 1.5 ± 0.5 pMC Tufa from old barrier. Collected October 1990 and submitted June 1991 by Lj Marjanac, INA- Inzenjering, Zagreb. 206 NHorvatincic et al.

Comment: (LjM) Tufa barrier covered by younger river and lake sediments. pMC Z-2552. Topoljski Buk 3 47.9 ± 0.8 Tufa from borehole in barrier. Collected by A Renic, Institute of Geology, Zagreb. 7-2653. Topoljski Buk 4 2.6 ± 0.8 pMC S13C -9.2%0 Hard, crystalline tufa from outcrop near road, close to Topoljski Buk waterfall. = pMC Z-2654. Topoljski Buk 5 59.3 ± 1.1 S13C = -9.6%0

Hard tufa from the top of dry barrier, right side of Topoljski Buk waterfall, inner part of sample. pMC Z-2673. Topoljski Buk 6 60.2 ± 1.2 b13C -9.1%0 Same as above, middle part of sample. = Z-2674. Topoljski Buk 7 60.3 ± 1.2 pMC S13C -9.5%o Same as Z-2654, outer part of sample. = Topoljski Buk 8 77.6 ± 1.3 pMC Z-2655. b13C Recent tufa from Topoljski Buk waterfall. = -9.7%o Z-2656. Topoljski Buk 9 46.0 ± 1.0 pMC b13C Hard tufa, near Z-2654, = -8.8%0 Z-2657. Topoljski Buk 10 20.7 ± 0.1 pMC b13C -7.3%o Hard tufa, ca. 5 m bellow Z-2654. = Z-2658. Topoljski Buk 11 74.0±1.3pMC S13C Hard tufa from the top of dry barrier. = Buk 12 61.0 ± 1.2 pMC Z-2659. Topoljski 13C Hard tufa, ca. 5 m bellow Z-2656. = Z-2660. Topoljski Buk 13 58.8 ± 1.1 pMC S13C Hard tufa near Z-2654. = -8.9%0

C) Topolje Quarry Series Huge deposits of old tufa near Knin, more than 20 m thick, formerly exploited as building material.

Z-1310. Topolje Quarry 1 1.4 ± 0.8 pMC Compact, layered tufa, west side of quarry. 230ThP 34U age 251 +291-24 ka. o13C = Z-1312. Topolje Quarry 2 0.4 pMC Compact tufa, middle layer, east side of quarry. Z-1314. Topolje Quarry 3 8.5 ± 0.3 pMC Porous tufa with moss structure (Cratoneurum commutatum), top of quarry. Z-1315. Topolje Quarry 4 3.7 ± 0.4 pMC Porous tufa with moss structure, same location as above. Inner part of sample. Z-2441. Topolje Quarry 5 5.0 ± 0.4 pMC Same as above, outer part of sample. Z-2597. Topolje Quarry 6 1.4 ± 0.8 pMC b13C Compact layered tufa from the west side of the quarry (cf. Z-1310). = -9.1%o Rudjer Boskovic Measurements XIV 207

Z-1318. Butinica R. 46.6 ± 0.7 pMC Butisnica River, a tributary of the Krka River, Golubic village. Compact tufa eroded at the surface, bridge near hydroelectric power plant.

D) Corica Buk Series

Z-2325. Corica Buk 1 78.9 ± 1.2 pMC Tufa from Corica Buk waterfall, 15.5 m high, artificial lake Brljan. Z-2326. (iorica Buk 2 80.7 ± 1.2 pMC Tufa from Corica Buk waterfall. S13C = -7.0%0 E) Manojlovac Series

Recent tufa and wood samples were collected from the Krka River banks and dry riverbed to determine the initial 14C activity of tufa. A stretch of Krka River with several waterfalls and primitive watermills has remained dry since the construction of hydroelectric plant Manojlovac in 1907 (Friganovic 1984), enabling us to collect samples that are not contaminated by bomb 14C. Z-2327. Manojlovac 1 74.1 t 0.8 pMC Recent tufa from 32-m-high tufa barrier, Manojlovac waterfall

Z-2328. Manojlovac 2 79.0 ± 1.5 pMC Crystalline tufa from barrier above waterfall 6'3C = -8.0%o Z-2329. Manojlovac 3 90 ± 120 Wood from Manojlovac mill. 99.1 ± 1.4 pMC Z-2330. Manojlovac 4 82.0 ± 1.3 pMC Recent porous tufa with moss structure, near Manojlovic mill, b13C = -7.2%0 Z-2397. Manojlovac 5 0.8 pMC Recent porous dry tufa from Manojlovac waterfall near Kevica mill.

F) Rosnjak Series

Z-2395. Rosnjak 1 79.5 ± 1.4 pMC Tufa from river bed, Krka River between Rosnjak and Miljacka waterfalls. b13C = -8.2%0 Z-2396. Roanj ak 2 80.6 t 0.7 pM C Tufa from dry riverbed, Krka River, upstream from Rosnjak waterfall.

Z-2617. Roki slap waterfall 90.6 ± 1.0 pMC Recent tufa from Roski slap waterfall, 15 m high. Coll A. Plenkovic, Faculty of Natural Sciences, Zagreb.

G) Kalica Kuk Series

Old tufa barrier between Lakes Visovac and Mlinarsko, presently 10-15 m above the lake surface. Comment: (NH) '4C content indicates the degree of contamination with recent carbon. Z-2647. Kalica Kuk 1 5.5 ± 0.6 pMC Hard, porous tufa, top of the barrier. b13C = -8.0%0 208 N Horvatincic et al.

Z-2648. Kalica Kuk 2 8.4 ± 0.6 pMC &'3C Hard, compact tufa, bottom of the barrier, from short tunnel dug into the barrier. = -8.4%0 Z-2649. Kalica Kuk 3 9.3 ± 0.8 pMC b13C Hard, compact tufa, top of the barrier. 230Th/234U age 113 ± 4 ka. = -9.7%o Z-2650. Kalica Kuk 4 2.4 ± 0.6 pMC b13C Hard, porous tufa, same location as Z-2648. 230Th/234U age 120 ±9 ka. = -8.9%0 Z-2651. Kalica Kuk 5 2.1 ± 0.8 pMC b13C Hard, compact tufa, same location as above. = -8.3%0 Z-2652. Kalica Kuk 6 1.4 ± 0.6 pMC b13C Layered tufa, same location as Z-2648. 230Th/234U age 116 ± 6 ka. = -8.7%0

H) Skradinski Buk Waterfalls Series

The waterfalls on Krka River with the greatest volume of water, consisting of 17 cascades, up to 100 m wide and 400 m of total length. Total height 45.7 m (Friganovic 1984).

Z-2661. Skradinski Buk 1 81.4 ± 1.3 pMC b13C Compact tufa, partly porous, from dry barrier, left bank of the Krka River. = -7.7%o Z-2662. Skradinski Buk 2 94.9 ± 1.4 pMC b13C Recent, porous tufa covered by moss, below waterfalls, left bank of Krka River. = Z-2663. Skradinski Buk 3 83.6 ± 1.3 pMC b13C Hard, porous tufa from dry barrier, above Z-2661. = Z-2664. Skradinski Buk 4 73.8 ± 1.3 pMC b13C Compact tufa from dry barrier, above Z-2663. = -9.7%o Z-2665. Skradinski Buk 5 78.2 ± 1.3 pMC Recent tufa in tubular form, right bank of the Krka River. b13C = -9.4%o Z-2618. Skradinski Buk 6 80.2 ± 1.3 pMC Recent tufa below waterfall, left bank of the Krka River. Collected in 1996 by A Plenkovic.

1) Lake Prukljan Series Submerged tufa barrier in Lake Prukljan formed by the Krka River. Collected in 1991 by D Petricioli

Z-2366. Lake Prukljan 1 105.3 ± 0.8 pMC Tufa from submerged barrier, water depth 3.5 m. Z-2367. Lake Prukljan 2 103.9 ± 0.7 pMC Shells separated from tufa Z-2366. Plitvice National Park Series

A) Lake Prosce-Matica River Series Sediment from Lake Prosce at Matica River mouth, 9 m water depth. Collected in August 1990 by D Petricioli and submitted by D Srdoc and N Horvatinc:ic. Rudjer Boskovic Measurements XIV 209

Z-2289. Lake Prone -Matica 1 96.2 ± 1.3 pMC Sawdust from watermills mixed with lake sediment, 0-6 cm depth. Z-2290. Lake Prone - Matica 2 13 pMC Same as above, 6-11 cm depth.

B) Lake Prosce Series

Analyses of shallow sediment cores from Lake Prosce to determine the increase of 14C activity in lake sediment by nuclear bomb effect and local anthropogenic pollution in the uppermost 30 cm thick layers (Srdoc et al. 1992a). The sediments were cored at two different water depths: 20 m and 32 m. Collected Mar, and Aug. 1990 by D Petricioli and submitted by D Srdoc and N Horvatincic. 14C Comment: Initial activity of Lake Prosce sediments is 72.0 pMC (Krajcar Bronic et al. 1992). Upper 5-10 cm of sediment was formed after ca. AD 1950.

Table 3 Lake Prosce sediment core Depth S13C Sample nr (cm) (%o) 20 m water depth Z-2222 0-5 ±0.6 Z-2223 5-10 72.4 ± 1.0 Z-2224 10-15 69.8 ± 1.1 Z-2225 15-20 66.3 ± 1.0 Z-2226 20-23 67.5 ± 1.0 32 m water depth Z-2237 1-5 1.1 Z-2275 5-10 -9.47 ±0.8 Z-2277 10-15 -9.31 ±0.7 Z-2276 15-20 -9.03 1.1 Z-2278 20-25 -8.97 1.0 Z-2279 25-29 -9.02 0.7

Z-2322. Lake Prone 136.8 ± 1.3 pMC '3C = -33.50%0 Organic residue of sediment after chemical pretreatment of sample Z-2222, 0-5 cm depth. Comment: Precipitation after AD 1950 is confirmed (Srdo et al. 1992a).

C) Lake Kozjak Series

Samples from recent sediment cores from Lake Kozjak to determine the increase of 14C activity in lake sediment by nuclear bomb effect, and local anthropogenic pollution in the uppermost 20-cm- thick layers (Srdoc et al. 1992a). Sediments were cored at water depths 21.5 m and 38 m in March and in August 1990 by D Petricioli. Previous measurements: Z-2116 to -2120 (Srdoc et al. 1992b: 170-171).

14C Comment: Initial activity of Lake Kozjak is 76.0 pMC (Krajcar Bronic et al. 1992). Upper 5- 10 cm of sediment precipitated after ca. AD 1950. 210 N Horvatincic et al.

Table 4 Lake Kozjak sediment core Depth b13C Sample nr (cm) (%o PDB) Core #1, 21.5 m water depth, March 1990 Z-2233 0-5 88.6±0.8 Z-2234 5-10 74.7 ± 0.7 Z-2235 10-15 72.6± 0.7 Z-2236 15-20 71.4 ± 0.5 Core #2, 38 m water depth, August 1990 Z-2266 1-5 -9.02 Z-2267 5-10 -8.88 ±0.7 Z-2288 15-20 -9.11 1.1 Core #3, 21.5 m water depth, August 1990 Z-2272 1-5 88.2 ± 0.9 Z-2268 5-10 76.9 ± 1.1 Z-2271 10-15 74.3 ± 1.1 Z-2269 15-19 -9.08 1.1 Z-2270 19-23 -9.11 Core #4, 21.5 m water depth, August 1990 Z-2295 0-1 -8.69 1.1 Z-2298 1-2 -8.90 1.2 Z-2299 2-3 91.0 ± 1.2 Z-2305 3-4 -9.03 1.1 Z-2310 4-5 89.7 ± 1.1 Z-2312 5-7 86.1±1.1 Z-2313 7-9 81.9±0.8 Z-2314 9-11 -8.92 1.1 Z-2315 11-13 77.0± 1.1 Z-2320 21-23.5 -9.16 1.1 Core #5, 21.5 m water depth, August 1990 Z-2296 0-1 80.0 ± 1.2 Z-2297 1-2 85.4 ± 1.1 Z-2304 2-4 81.1± 1.1 Z-2311 4-5 82.4 ± 1.1

TREE RINGS Hungary, Mt Matra Cellulose from spruce tree rings (1956-1986) (Picea sp.), Mt Matra, northern Hungary (47°54'N, 19°55'E), 650 m asl. Collected and submitted in July 1987 by K Kozak, Institute of Isotopes, Hun- garian Academy of Science, Budapest, Hungary. Comment: (KK) The tree was felled at the end of the 1986 growing season. Cellulose preparation is 14C described in Kozak et al. (1989). Comparison with other atmospheric and tree-ring data in Kra- jcar Bronic et al. (1998). Rudjer Boskovic Measurements XIV 211

Table 5 Spruce tree rings, Mt Matra Sample nr Year a14C (%o) Z-2012 1956 63.8± 9.6 Z-2013 1959 279.6± 14.0 Z-2014 1961 209.2±12.3 Z-2015 1963 662.6 ±20.1 Z-2049 1964 869.1 ± 19.4 Z-2048 1965 750.0± 19.2 Z-2016 1966 700.7 ± 18.8 Z-2017 1969 560.0 ± 17.5 Z-2018 1972 462.5 ± 14.4 Z-2019 1975 426.1 ± 14.4 Z-2020 1978 368.2 ± 13.8 Z-2021 1981 247.6 ± 12.9 Z-2022 1984 280.2 ± 12.8 Z-2023 1986 190.4 ± 12.3

Croatia, Plitvice Lakes National Park

Cellulose from tree rings (1960-1986) of spruce (Picea sp.) from Plitvice National Park. The tree was felled in summer 1987. Collected and submitted 1987 by B Obelic and K Kozak. '4C Comment: activity in tree rings reflects global atmospheric variation not affected by local industrial CO2 emissions. 14C activities of tree rings are in good agreement with 14C activities of atmospheric CO2 (Obelic et al. 1992; Krajcar Bronic et al. 1998) and Mt Matra tree ring 14C activities. Slovenia, Krko

Linden (Tilia sp.) tree rings (1980-1983), from Libna, near Krsko (45°55'N, 15°31'E), ca. 180 m asl, ca.1 km east of the nuclear power plant Krsko. The tree was felled in 1984.

Comment: 14C activities of tree rings are in good agreement with the 14C activities of atmospheric CO2 (Obelic et al. 1992; Krajcar Bronic et al. 1998), as well as with the Mt Matra and Plitvice tree- ring 14C activities.

ACKNOWLEDGMENTS

We thank E Hernaus and D Voscak for sample preparation. We also thank MA Geyh, Geowissen- schaftliche Gemeinschaftsaufgaben, Hannover, Germany, and J Pezdic, Jozef Stefan Institute, Ljubljana, Slovenia for stable isotope measurements. 212 N Horvatincic et al.

Table 6 Plitvice tree rings Sample nr Year A14C (%o) Z-2375 1950 9.1 9.7 Z-2380 1960 352.3 12.9 Z-2194 1961 238.2 12.3 Z-2195 1963 762.6 19.1 Z-2372 1963 756.5 15.7 Z-2196 1964 875.2 18.7 Z-2377 1965 760.9 13.6 Z-2197 1965 738.5 18.9 Z-2198 1966 699.7 16.2 Z-2379 1967 739.1 16.0 Z-2199 1968 591.9 17.4 Z-2394 1969 557.4 18.3 Z-2393 1971 155.7 19.2 Z-2200 1972 497.0 19.9 Z-2376 1973 474.8 10.4 Z-2392 1974 433.8 17.2 Z-2201 1975 399.0 15.2 Z-2402 1976 380.5 16.8 Z-2378 1977 478.3 12.7 Z-2218 1978 134.7 21.4 Z-2373 1979 341.4 11.0 Z-2219 1981 278.0 11.8 Z-2220 1984 238.4 13.7 Z-2221 1986 196.5 19.2 Z-2374 1986 223.3± 9.8

Table 7 Libna near Krsko tree rings Sample nr Year l4C (%o) Z-1413 1980 360.7±15.2 Z-1412 1981 258.1 ± 14.8 Z-1411 1982 245.7± 14.7 Z-1410 1983 222.8 ± 14.2 Z-1408 Bark (outer layer) 188.0 ± 13.9 Z-1409 Sapwood 325.3± 14.6

REFERENCES 14C Babic Lj, Lackovic D, Horvatincic N. 1996. Meteoric Kozak K, Obelic B, Horvatincic N. 1989. Tritium and phreatic speleothems and the development of cave in the tree rings of the last three decades. Radiocarbon stratigraphy: an example from Tounj Cave, Dinarides, 31(3):766-70. Croatia. Quaternary Science Reviews 15:1013-22. Krajcar Bronic I, Horvatincic N, Obelic B. 1998. Two de- Friganovic M. 1984. Krka-od antickog Titiusa do nacio- cades of environmental isotope record in Croatia: re- nalnog parka (in Croatian). Turisticke Monografije construction of the past and prediction of future levels. Broj 4. Zagreb: Privredni Vjesnik. 124 p. Radiocarbon 40(1):399-416. Horvatincic N, Bistrovic R, Obelic B. 1996. Radiocarbon Krajcar Bronic I, Horvatincic N, Srdoc D, Obelic B. 14C and uranium-series dating of travertine. Acta Geolog- 1992. Experimental determination of the initial ica Hungarica 39:77-80. activity of calcareous deposits. Radiocarbon 34(3): Rudjer Boskovic Measurements XIV 213

593-601. Srdoc D, Horvatincic N, Krajcar Bronic I, Obelic B, Obelic B. 1989. The Radiocarbon Data Base at Rudjer Sliepcevic A. 1992b. Rudjer Boskovic Institute radio- Boskovic Institute Radiocarbon Laboratory. Radio- carbon measurements XII. Radiocarbon 34(3):155- carbon 31(3):957-1062. 75. Obelic B, N, Horvatincic Krajcar Bronic I.199214C con- Srdoc D, Horvatincic N, Obelic B, Krajcar Bronic I, centration in tree rings in the Plitvice National Park Sliepcevic A. 1987. Rudjer Boskovic Institute radio- area. In: Proceedings of the 1st symposium of Croat- carbon measurements IX. Radiocarbon 29(1):115-34. ian Radiation Protection Association Zagreb, Croatia; Srdoc D, Obelic B, Horvatincic N, Krajcar Bronic I, 24-26 Nov 1992. p 247-50. Marcenko E, Merkt I, Wong HK, Sliepcevic A. 1986. B, Obelic Horvatincic N, Srdoc D, Krajcar Bronic I, Radiocarbon dating of lake sediment from two karst Sliepcevic A, Grgic S. 1994. Rudjer Boskovic Insti- lakes in Yugoslavia. Radiocarbon 28(2A):495-502. tute radiocarbon measurements XIII. Radiocarbon Srdoc D, Obelic B, Horvatincic N, Krajcar Bronic I, 36(2):303-24. Sliepcevic A. 1984. Rudjer Boskovic Institute radio- Obelic B, Smalcelj M, Horvatincic N, Bistrovic R, carbon measurements VIII. Radiocarbon 26(3):449- Sliepcevic A. 1995. Radiocarbon dating of the Zagreb 60. upper town prehistoric settlement. Radiocarbon Srdoc D, Obelic B, Horvatincic N, Krajcar Bronic I, 37(2):259-66. Sliepcevic A. 1989. Rudjer Boskovic Institute radio- Ramsey CB. 1995. Radiocarbon calibration and analysis carbon measurements XI. Radiocarbon 31(1):85-98. of stratigraphy: the OxCal program. Radiocarbon Srdoc D, Osmond J, Horvatincic N, Dabous A, Obelic B. 37(2):425-30. 1994. Radiocarbon and uranium series dating of the Ramsey CB. 1998. The OxCal Program Manual, v.3.0. Plitvice Lakes travertines. Radiocarbon 36(2):203- Website http://www.rlaha.ox.ac.uk/oxcal/oxcal.htm. 20. D, Srdoc Breyer B, Sliepcevic A. 1971. Rudjer Boskovic Srdoc D, Sliepcevic A, Obelic B, Horvatincic N. 1979. Institute radiocarbon measurements I. Radiocarbon Rudjer Boskovic Institute radiocarbon measurements 13(1):135-40. V. Radiocarbon 21(1):131-7. D, Srdoc Horvatincic N, Ahel M, Geiger W, Schaffner Srdoc D, Sliepcevic A, Planinic J, Obelic B, Breyer B. Ch, Krajcar Bronic I, Petricioli D, Pezdic J, Marcenko 1973. Rudjer Boskovic Institute radiocarbon measure- E, Plenkovic-Moraj A. 1992a. Anthropogenic influ- ments II. Radiocarbon 15(2):435-41. 14C ence on the activity and other constituents of re- Stuiver M, Polach HA. 1977. Discussion: reporting of cent lake sediments: a case study. Radiocarbon 34(3): 14C data. Radiocarbon 19(3): 355-63. 585-92.

RADIOCARBON CALIBRATION BY THE DATE DISTRIBUTION METHOD

Paul Muzikar

Department of Physics, Purdue University, West Lafayette IN 47907, USA

ABSTRACT. A method is presented for calibrating radiocarbon ages based on statistical analysis of a large number of randomly distributed dates. One interesting feature of this method is that it is internal; that is, it allows one to extend a known calibration curve further 14C back in time by using only dates, with no reference to any other dating technique. A serious difficulty in implementing this method lies in assembling a sample of dates with the correct statistical properties. INTRODUCTION

Great success has been achieved over the past few decades in calibrating the radiocarbon time scale. Tree rings have provided a very accurate calibration going back about 12,000 yr, while the U-Th dating of corals has been used to extend the calibration back to around 30,000 yr. For an up-to-date discussion of this progress, the reader is referred to Stuiver et al. (1998), along with accompanying articles in the same issue of Radiocarbon. The calibration effort has an important influence on many fields of research; thus, it is fitting to bring to bear on this effort as many independent lines of thought as possible.

My purpose in this paper is to discuss a somewhat speculative method for calibrating the 14C time scale, referred to here as the Date Distribution (DD) Method. This method, as the discussion will make clear, offers only a limited resolution in time. However, it has the interesting feature of being an internal calibration method. That is, once a calibration curve is established, the DD method can 14C extend it back in time by using only dates, with no necessity of using another dating method. The word "speculative" was used in the preceding paragraph for a particular reason. A serious difficulty would have to be overcome in applying the DD method. As explained in the next section, a set 14C of dates satisfying certain criteria must be assembled in order for the DD method to work. At present, it is not clear exactly how this should be done, although it is by no means demonstrably impossible. To some extent, then, this paper constitutes a "thought experiment".

In the next section I explain the DD Method, and try to highlight the various difficulties that could arise in implementing it. In the following section I present a numerical experiment illustrating how the method works. This experiment also allows us to discuss the level of precision that could be expected in any real effort to implement the method.

METHOD

To examine the DD method in some detail, we must first establish our notation. We will plot dates in the X-Y plane, with x denoting the true date (in cal BP) and y denoting the conventional 14C age (in BP). A calibration curve relates these two dates, and is given by

y = q(x) (1)

The goal of calibration research is to determine the function fi(x). Now, imagine that on a certain interval a S x < b, we have M dates distributed randomly on the X- axis; that is, each date is equally likely to have any value between a and b. Thus, the probability density for a given true date is given by:

215 216 P Muzikar

1 (x) (2) Pi b-a

measure y, the con- for x between a and b, and is zero otherwise. For each of these dates, we actually thus investigate the prob- ventional age. The values of y will lie between ya = q(a) and yb = q(b). We intervals dx and dy we ability density for y, which we will call p2(y). Since for any corresponding must have

(3) pi (x)dx = P2 (Y)dY

we can t h en say th at

p dx p2

Or, with our particular form for pl(x),

'(x) = ( b-a )p2(Y) distribu- The idea of the DD method is based on Equation (5). Here is how it would work. From the an tion of the actual, measured uncalibrated dates lying between ya and yb, one would construct Equation (5) to deter- approximation to the function p2(Y). However, now comes a key point. To use we only mine the slope of the calibration curve, we need to know the value of (b-a); from the data recent portion of our measured conventional know ya and yb. To overcome this problem, we let the to a ages overlap with a known calibration curve. Thus, we determine (b-a) by fitting Equation (5) known portion of the calibration curve. the cal- Note that Equation (5) only gives us the slope of the calibration curve. So, to use it to extend Thus, in ibration back in time, we need at least one point from an already existing calibration curve. applying the DD method, we need to work from an already existing calibration curve for 2 reasons: because the value of (b-a) is not generated by the data because the method only gives us the first derivative of the calibration curve.

To summarize, here are the basic steps involved in implementing the DD method: 14C have a flat prob- 1. Collect a large number of ages y on an interval ya to yb. These events should ability distribution for their true ages. 2. Use this data to construct an approximation to the distribution p2(y). (b-a). 3. Fit Equation (5) to a known portion of the calibration curve to determine the value of an 4. Use Equation (5) to find the slope of the calibration curve, and to extend the curve back into unknown region of time.

Up to this point in this section we have glossed over the key difficulty (and perhaps the most interesting philosophical question) concerning the DD method. For the method to work, we must use a large sample of events whose true dates are distributed with even probability over some particular time interval. Clearly, much archaeological or geological insight is required in assembling such a sample. For example, if a sample had many dates clustered around a particular value, simply because that value was a heavily dated, interesting archaeological episode, it would be an unsuitable sample. Calibration by the Date Distribution Method 217

Perhaps an appropriate group of dates would be a set of 14C dates for various geological phenomena from all over the world. In any event, the compilation of such a sample is a question worthy of comment from knowledgeable researchers.

One other small point can be made. We have presented our discussion assuming that the function fi(x) has a positive slope for all values of x. This is not necessarily true for realistic calibration curves. The DD method, however, will construct a quite coarse-grained version of the function, where many short-lived features are washed out. This coarse-grained function is much less likely to have a negative slope at any points.

NUMERICAL EXPERIMENT

To show how the DD method would be used in practice, let us now work through a numerical experiment. For the sake of illustration, we use a hypothetical calibration curve, a polynomial generated to have many of the features of the actual calibration curve presented by Stuiver et al. (1998). We consider the time interval, in real years, of 10,000-18,000 cal BP. For our hypothetical calibration curve, this corresponds to an interval in uncalibrated years of 9000-15,000 BP. The functional form of our curve is given by the following polynomial:

fi(x) _ .1312307 x 10-6 x7 - .3256363 x 10-5x6 - .0001273x5 +.00581797x4 - .0773117x3 (6)

+.303132x2 +1.92271x-6.7332 .

Here, x and y are measured in units of ka. Figure 1 shows the calibration curve, while Figure 2 shows a plot of its first derivative. We have selected a function that has reasonable deviations from linearity, in that its slope varies from its average value by about 20%. Larger deviations from linearity would be easier to see with our statistical method, while smaller deviations would be more difficult to see. We conduct our experiment in several steps. We first use a random number generator to produce 2400 dates randomly distributed between 10,000 cal BP and 18,000 cal BP. We then use our hypothetical calibration curve to turn these into 2400 uncalibrated dates lying between 9000 BP and 15,000 BP. Because the calibration curve is not linear, these uncalibrated dates are not evenly distributed between the two limits. They will tend to be bunched into regions where the calibration curve has a lower slope. This, of course, is the basis of the DD method.

The idea of the experiment is to take those 2400 uncalibrated dates as our data, and then try to recon- struct the calibration curve from them. There are many particular ways to use Equation (5) to do this reconstruction. Here, we will use a simple, straightforward method. We divide up the uncalibrated period into 6 equal, 1000-yr intervals and then we count how many of the events are in each interval; Table 1 shows these numbers. From these numbers we obtain our experimental determination of pi, the probability of an event being in 1 of the 6 intervals. For each of the 6 intervals, we estimate the average slope si for that interval as

(7) where A is a constant. We thus are approximating the calibration curve as a series of line segments with slopes si. The constant A is determined by matching the average slope in the first interval (9000-10,000 BP) to the known value from the calibration curve. This determination of the value of A is equivalent to the determination of the value of (b-a), a procedure discussed in the previous sec- 218 P Muzikar

15d

14+

19d

114-

14C Figure 1 calibration curve given by Equa- ,o+ tion (6). The X-axis is the true, calibrated date, while the Y-axis is the uncalibrated 14C date. Both dates are measured in ka.

s+ 14 cal BP (ka)

0.e+

0.85

0.75

Figure 2 The first derivative, or slope, of the cal- 0.7+ ibration curve given by Equation (6), plotted as a function of the true age in ka. Note that the slope is dimensionless.

10 1 14 16 18 cal BP (ka)

tion. In Table 1 we also give the expected value of the average slope for each interval; this expected value is computed, using the calibration curve, from the oy and the Ox of each interval by the simple relation S.. oy (8) A Calibration by the Date Distribution Method 219

Table 1 Results of the numerical experiment with 2400 events. The number of events in each time interval is given; this number is divided by 2400 to give pi. The experimentally determined average slope for each interval is then given by si, whereas the expected average slope (computed from the calibration curve) is given in the last column. No value for si is predicted in the first time interval, since the constant A was determined by using the pi for this interval. The accuracy of the method can be judged by comparing the numbers in the last 2 columns. Number of Interval events. si 9-10 ka 430 .179 10-11 ka 432 .180 11-12 ka 412 .172 12-13 ka 376 .157 13-14 ka 370 .154 14-15 ka 380 .158

Table 2 Results of the numerical experiment with 4800 events. Notation is the same as in Table 1, except that p; is computed by dividing the number of events by 4800. Number of Interval events s; 9-10 ka 901 .188 10-11 ka 860 .179 11-12 ka 788 .164 12-13 ka 765 .159 13-14 ka 782 .163 14-15 ka 704 .147

Table 2 shows results for a larger sample of 4800 randomly generated events. The crucial check on the method is a comparison of the last 2 columns in the tables, which contain the experimentally estimated average slope and the actual average slope. We can see that in Table 1 the trend of an increasing slope is roughly captured by the data. The results in Table 2 are noticeably better.

We can now discuss random fluctuations, and how they affect the accuracy of this method. If M is total number of points (in our case 2400 or 4800) and Ni is the average number of points expected in a given interval (in our case 1 of our 6 intervals), then it is easy to show that

M (9)

Here ANi is the variance in N. Thus, to a good approximation we can say that

4Ni 2 Thus, we can say that

p1 (11) dpi - i (Ni) 2 and 220 P Muzikar

(2)2 ~ si (12) As _ . (N)2 the The i2 factor in the last equation appears when we take into account, in an approximate manner, So, for additional variance due to the uncertainty in the value of the experimentally determined A. the 2400-event sample, we estimate that the uncertainty in slope determination is roughly As, in 0.049, while for the 4800-event sample we estimate As, 0.035. So, for a given number of events our sample, a certain resolution in slope determination is expected.

CONCLUSION the 14C We have discussed the general principles behind the Date Distribution method for calibrating time scale, and worked through a numerical experiment to show how it could work in practice. It is 14C clear that a crucial difficulty in applying the method lies in assembling a sample set of dates with the correct statistical properties. If the method seems promising enough for further effort, we hope archaeological and geological insight can be brought to bear in assembling such a sample.

ACKNOWLEDGMENTS

I am grateful to David Elmore, Nick Giordano, and Darryl Granger for useful discussions.

REFERENCE G, Plicht J, Spurk M. 1998 Stuiver M., Reimer PJ, Bard E, Beck JW, Burr GS, Hughen KA, Kromer B, McCormac van der INTCAL98 radiocarbon age calibration, 24,000-0 cal BP. Radiocarbon 40:1041-77. LETTER TO THE EDITOR

BIAS, ACCURACY, AND PRECISION 2 July 1999

Dear Sir,

In the most recent issue of Radiocarbon (Vol 41, Nr 1, 1999), a paper by Rasmussen et al. discusses a blind check of the accuracy of the Copenhagen radiocarbon dating system. The authors report on 14C a comparison of a series of dates from dendrodated samples, measured over a 23-yr period, both with the absolute dendrodates and also with the bidecadal calibration curve (Stuiver and Pearson, 1993).

The authors compare 92 uncalibrated 14C dates of the Copenhagen measurements and the equivalent 14C uncalibrated dates from the Stuiver and Pearson bidecadal calibration curve (1993). The comparison is based on the difference between each pair of measurements and its error (calculated as the square root of the sum of the squared counting errors for each result). From the results, they calculate the average of the differences (or if necessary the weighted average) as 54 yr and present the histogram of the individual differences (see their Figure 2) as well as quantifying the variation in the results by the standard deviation (quoted as ± 72 yr). Given the symmetry of the histogram, it seems reasonable to assume that the underlying distribution of differences is Normally distributed so that 95% of the individual differences should lie in the range 54 ± 2 x 72, or -90 to 200 yr. This is borne out in the histogram. The authors then conclude "the comparison shows a good agreement, well within 16 between the 14C measurements" and that "good accuracy can be obtained".

I would like to draw the authors' attention to a paper which deals with the comparison of two sets of measurements (Bland and Altman 1986). Bland and Altman suggest that figures such as Figure 1 of Rasmussen et al. can be difficult to interpret and may be misleading, and that to measure agreement, it is useful to plot the difference between the two measurements against the average of the two measurements. In this case, the summary of agreement is the mean difference (or "bias") and the standard deviation of the differences. The precision of the agreement is then quantified by the estimated standard error of the mean difference. Bland and Altman avoid the use of the word "accuracy".

Rasmussen et al. only briefly mention laboratory bias, preferring to use the term accuracy, but it is worth digressing briefly to considering accuracy, bias, and precision, since the definitions of these terms are relevant to the discussion of this paper.

An accurate result is one which is close to the true value and which is precise. An accurate result should have zero bias, where bias in the estimation sense is defined as the difference between the expected value of the statistic (or parameter estimate) and the true value of the population parameter. Here, the authors are trying to estimate the true difference between the Copenhagen and the calibration results, which is estimated by the mean of the differences between the two sets of dates. Preci- sion refers to how varied the measured results are, and so it measures the spread or scatter in the results, but in an estimation sense, precision is the error on the estimate, which in this case would be the error on the mean difference.

Having calculated the mean difference and standard deviation, the authors seem to rest their case concerning the accuracy of the Copenhagen results, since 54 ± 72 yr shows good agreement ("well

221 222 Letter to the Editor

agreement within ± 16") with the calibration data and the quoted standard deviation is in reasonable with the estimated value. In fact, they have summarised the agreement between the two laboratories. However, we need to be careful at this point: if we are considering the accuracy of the laboratory, and then 54 yr is the estimate of the true difference between the Copenhagen and calibration results, precision of the 72 yr is a measure of the variation within the population of differences; but the by the agreement depends on the estimated standard error, which is the standard deviation divided the square root of the number of samples. Thus the precision with which we are able to estimate mean difference is 7.5 yr (72 divided by 9.6). with Thus, from the differences, the best estimate of laboratory agreement is 54 yr, and the precision which we have estimated the agreement is 7.5 yr. Together, a plausible range of values for the laboratory agreement can be estimated, or in other words, a 95% confidence interval can be calculated the value as 54± 15 yr (26), or 39-69 yr. This interval is highly significant, since it does not include 0; we can conclude that the Copenhagen and Stuiver and Pearson results are on average different (i.e., do not agree), with the Copenhagen results highly likely to be between 39 and 69 yr younger than the Stuiver and Pearson results. Thus I would dispute the description of the results as being accurate. There is, of course, the question of the validity of this calculation: the samples are not identical, they may not have exactly the same true age, and the measurements were made over a period of 23 yr, so by the it is likely that laboratory conditions changed in that time. The use of the term "accuracy" authors implies the existence of a true value (and hence the possible existence of a bias), but in this case, the authors are really measuring agreement, so that this might have been a better term to use. These arguments aside, I believe that it is still valid to make this estimate of agreement and that we must calculate the precision of this estimate, irrespective of how it is called, using the appropriate term. Marian Scott Department of Statistics The University of Glasgow

REFERENCES dat- Bland JM, Altman DG. 1986. Statistical methods for as- check of accuracy of the Copenhagen radiocarbon sessing agreement between two methods of clinical ing system. Radiocarbon 41(1):9-15. bidecadal measurement. The Lancet 8476:307-10. Stuiver M, Pearson GW. 1993. High-precision 1950- Rasmussen KL, Tauber H, Bonde N, Christensen K, calibration of the radiocarbon time scale, AD 35(1):1-24. Theodorsson P. 1999. A 23-year retrospective blind 500 BC and 2500-6000 BC. Radiocarbon RADIOCARBON UPDATES

Internet Resources

Over 600014C dates from archaeological and vertebrate palaeontological sites in Canada composing the Canadian Archaeological Radiocarbon Database are now available on the World Wide Web in the form of two online initiatives.

The main searchable CARD database, compiled by Dr Richard E Morlan of the Canadian Museum of Civilization, is located at http:l/www.canadianarchaeology.com/radiocarbon/card/card.htm

In addition, Dr Morlan, Arthur S Dyke and Roger N McNeely have created a program called "Map- ping Ancient History" that uses interactive maps for visual display of the locations of the sites in the CARD database, along with an animated representation of the spread of archeological sites over time. The URL of this project is:

http:l/wwwims 1.gsc.nrcan.gc.ca/projects/mabasyquad/radiocarbon/indexC 14.htm Lab Relocation

The Desert Research Laboratory (DRI) is terminating its operations in Las Vegas, Nevada, and will be transferring much of its radiocarbon dating instrumentation to the existing 14C dating laboratory at the Center for Applied Isotope Studies, University of Georgia (Athens, Georgia, USA). DRI staff members will not be relocating with the lab. However, DRI lab head Herbert Haas will spend 2 months at the new location, assisting in the transition and putting together a joint publication on sample preparation. All correspondence regarding DRI should be sent to: Dr. Herbert Haas c/o Center for Applied Isotope Studies University of Georgia 120 Riverbend Road Athens, GA 30602-4702 USA

Retirement

DRI lab head Herbert Haas is retiring and has started his own consulting business, RC Consultants, Inc. in Las Vegas, Nevada. Services include pretreatment of field samples for conventional radiocarbon and AMS dating, as well as assistance in field collection and curating of chronometric samples, training laboratory staff, and assisting labs during temporary staff shortages. Send correspondence to: RC Consultants, Inc. 2846 Marida Ct. Las Vegas, NV 89120 USA Tel: +1 702 434 3642; Fax: +1 702 434 4966 Email: [email protected]

Laboratory Profile

The Consejo Superior de Investigaciones Cientificas (CSIC) Geochronology Laboratory, situated in the Instituto de Quimica Fisica "Rocasolano" (Madrid, Spain), was built to provide a radiocarbon dating service to the Spanish archaeological community in the 1970s. It started with a proportional

223 224 Radiocarbon Updates

scintillation spec- counter using CO2 as a counting gas. In the mid-1980s it bought an LKB liquid since then. trometer and a new line to synthesize benzene was built. Both lines have been working most of them from To date, the CSIC Geochronology Laboratory has dated about 1500 samples, and South Spanish archaeological sites, though a significant number of dates are from Portugal America (Argentina and Brazil). The laboratory abbreviation is CSIC. For further information, please contact: Dr Fernan Alonso and Mr Antonio Rubinos Geochronology Laboratory Instituto de Quimica-Fisica Rocasolano - CSIC Serrano,119 28006 Madrid, Spain Phone: +34 91 561 9400 Fax: +34 91 564 2431 Email: [email protected] or [email protected]

Faculty Appointment Dr W G Mook of the Groningen Centre for Isotope Research informs us that his faculty position has been appointed to Dr Harro A J Meijer. Meijer has a background in physics and a broad interest in der the applications of isotope studies. The department's scientific staff also include Dr Hans van Plicht and Dr Henk Visser. There is an open position for a 4th scientist.

Company Relocation Packard BioScience BV has expanded and relocated to a new building. Their new address is: Packard BioScience BV P.O. Box 5205 9700 GE Groningen, The Netherlands Phone: +31 50 544 59 00 Fax: +31 50 544 59 50 Email: [email protected] Website: http://www.packardinstrument.com CORRECTION

We have learned that erroneous numbers were printed in two cells of Table 1, p.1123, of the following article from INTCAL98, the recent updated calibration issue:

Kromer B, Spurk M. Revisions and tentative extension of the tree-ring based 14C calibration, 9200-11,855 cal BP. Radiocarbon 40(3):1117-25. In the "cal BC" column, the figure for HD-9264 should be 7891; the figure for HD-14088 should be 7886. A corrected version of the table appears on the following page.

A corrected version of the original page is also available as a PDF file on our website: http:l/www.radiocarbon.org/Journal/v4on3/page1123.pdf

It may be printed and then inserted or pasted over the page in question.

225 226 Correction

(Kromer and Becker TABLE 1. Revised German oak data. Compared to the previous publication are included 1993) data were added at the beginning of the chronology, and shifts of the dendroscale (see text).

Lab code Lab code BP BP (Hd-) cal BC BP BP (Hd-) cal BC 7871 9820 21 19079 8329 10278 14106 7851 9800 23 19055 8307 10256 24 14171 7836 9785 30 19054 8287 10236 27 9255 7831 9780 24 19091 8267 10216 23 14160 7821 9770 21 19175 8247 10196 25 14172 7717 9666 30 19174 8227 10176 23 8510 9658 30 19179 8207 10156 20 8511 7709 7701 9650 30 19349 8187 10136 20 8518 9643 30 19415 8147 10096 24 8519 7694 7679 9628 30 15145 8113 10062 23 8524 9618 30 14904 8108 10057 28 8525 7669 7659 9608 30 19350 8107 10056 20 8544 9602 30 15889 8103 10052 23 8141 7653 9586 33 15923 8098 10047 18 8140 7637 9569 30 16046 8068 10017 20 8091 7620 9562 30 15906 8053 10002 20 8144 7613 9562 30 16632 8043 9992 23 8286 7613 9552 30 16038 8033 9982 20 8145 7603 9543 30 16587 8023 9972 20 8295 7594 9537 30 16586 8018 9967 20 8127 7588 9530 30 16585 8013 9962 23 8244 7581 7578 9527 30 19351 8007 9956 24 8151 9519 30 9513 7968 9917 30 7758 7570 9507 30 14075 7961 9910 20 8171 7558 9490 30 9502 7956 9905 30 7759 7541 9472 30 14076 7951 9900 20 8273 7523 9422 30 9501 7946 9895 30 8304 7473 9409 30 14077 7941 9890 20 7760 7460 30 9492 7936 9885 30 8086 7423 9372 9370 30 14079 7931 9880 20 7757 7421 30 9491 7926 9875 30 30 14112 7921 9870 30 30 9486 7916 9865 30 30 16623 7911 9860 22 30 14274 7901 9850 23 30 16605 7896 9845 21 30 9264 7891 9840 30 30 14088 7886 9835 21 30 9258 7876 9825 30 `ANTIQUITY was founded in 192 - by de, pt tit h a s become the leading archaeological review - popular, authoritative., wltb a large and world-wide circulatiOh.

ITI I II[T 'S 2 Y old rep tat p 99PAnnes as the leading international jourr a of arcltaeotvgeal reporting and debate. The journal pu>iishes interesting, topical and c essible articles for awide and ence. Read by stu dents and amateurs, professors and pro- fessionals. Published four times ayear, ANTIQurTY and its SUPPLEMENT keeps you up to date with all the latest lhf©rmation on conferencesbooks, and archaeological events.

Papers range in time focus front Palaeolithic to present

1. 1 on allrtspa ofheto wrld reporting issues of interest and importance cart, *N 40thc ds & technologies New discoveries Heritage Issues & Museums Theory & Ethics ' Management & Landscapes Civilizations & Sites Reviews of all the latest books

Editor Deputy EdIt+otr Reviews Editor Caroline Malone Simon Stoddart Nicholas Jalne'

Contact us on all editorial and advertising matters at: Antiquity Office, New Hall, Cambridge C133 ft, Tel: (44)-1223-762298: Fax: (44)-1223-357075 e-mail: [email protected] http://intarch.ac.uk/an.tiquity

1999/2000 UK & USA & rest Euro rate Air Mail subscription rates: Europe of the world surcharge personal £35 US$66 55 $17/£10/i4 institutional £65 US$115 110 $20/£1 l/15 & a reduced *student rate: £23 (*applicable for two years after completion of course)

Subscriptions Department, The Company of Biologists, Bidder Building, 140 Cowley Road, Cambridge c134 4DL, England. Tel. (44) 1223.426164, Fax (44) 1223-423353 -mail: salesC?biologists.com Professional Analytical Services from: The Center for Applied Isotope Studies University of Georgia

RECOMMENDED QUANTITY: 1-5 g C for benzene LSC $225 < 1 g C forAMS $400 ("C/ "C included)

mg organic matter or 20 $40 100 mg carbonate

2mlH2Oor $50 100 mg carbonate

2 ml H2O $50 D/H 20 mg organic matter $70

1SN/ 14N 20 mg organic matter $50

iii LOW-LEVEL MDA to 20 pCi/L (7.5TU) $125 3 HuL LTRA MDA to 2 pCIIL (0.7 TU) O W-LEVEL (with tritium enrichment) $225

Quantity or Combination Discounts and Priority Service Available

CAIS (706) 542-1395 120 Riverbend Road Fax (706) 542-6106 Athens, GA USA 30602-4702 cais*,arches.uga.edu Jensa[en'i 200o D'iVFil'

17th International Radiocarbon Conference

It is our great pleasure to invite you to participate in the 17th International Radiocarbon Confer- ence, scheduled for June 18-23, 2000, in Israel. The Conference will be held at a beautiful location, in the rural setting of Kibbutz Ma'ale Hachamisha in the Judean Hills, just west of Jerusalem. The Kibbutz offers an attractive self- contained arrangement of excellent accommodation and conference facilities, which will enable a high degree of interaction between conference participants. The City of Jerusalem with its unique history and tourist attractions is a short drive away and is easily reached by bus or taxi. The first 14C Conference at the dawn of a new millennium will undoubtedly include exciting new scientific developments. Keeping the tradition of past Radiocarbon conferences, the scientific program will include a wide variety of topics. Sessions will be devoted to: *Archaeology- with a special session on 14C data of Global change historical periods in the Near East Glaciology Calibration of the 14C time scale Hydrology Sample treatment and measurement techniques Oceanography Geophysics and Geochemistry of 14C Geology Cosmogenic radionuclides Soils Environment past and present

Participants are welcome to indicate their preference for either oral or poster presentation of their papers. However, the final decision regarding form of presentation will be made by the Organizing Committee. Detailed information will be included in the Second Announcement. The social program of the Conference will include an afternoon walking tour in the Old City of Jerusalem and a one-day tour in the unique Dead Sea area (lowest point on the earth's surface). A floating swim in the Dead Sea - which is seven times saltier than the ocean - is indeed a unique fun experience. Sincerely, The Organizing Committee: Israel Carmi, Chairperson, The Weizmann Institute of Science Elisabetta Boaretto, Secretary, The Weizmann Institute of Science Hendrik J. Bruins, Ben Gurion University of the Negev Michael Paul, Hebrew University of Jerusalem Dror Segal, Hebrew Antiquities Authority Yoseph Yechieli, Geological Survey of Israel

For further information, please see the Conference website: http://www.radiocarbon.co.il/ Jerusa[en'i 200o D'5Vfl1' RADIOCARBON Back Issues Price Reduction-Save up to 87% Selected copies just $5 each. Conference proceedings $10.

Write in the number of copies Calculate Payment desired for each issue

YEAR VOLUME NR 1 Regular issues: copies x$5ea.=$ 1980 22 issues: 1981 23 *Conference Proceedings copies x $10 ea. = $ 1982 24

1983 25 Subtotal: $

1984 26 Shipping via surface mail: 1985 27 (Contact us for other shipping methods) Add ea. book for US 1986 28 $2 Add $3 ea. book outside US 1987 29 copies x $ ea. = $ 1988 30

1989 31 Shipping total: $ 1990 32 Total due: $ 1991 33

1992 34

Payment & Shipping Address MasterCard Visa Check enclosed Please bill me

Name:

Address:

Fax: E-mail:

For Credit-Card Orders

Card number: Expiration date:

Signature: Phone:

For online Contents see: http: //www. radiocarbon. org/Pubs/contents . html

Prices good while supplies last. Mail or fax this form to: RADIOCARBON, 4717 East Fort Lowell Road, Room 104, Tucson, Arizona 85712-1201 USA Phone: +1 520 881-0857; Fax: +1 520 881-0554; [email protected] 14. RADIOCARBON The University of Arizona Department of Geosciences ( An International Journal of c 4717 E. Ft. Lowell 6010 Cosmnogenic Isotope Research Rd., Rm. 104 Tucson, AZ 85712-1201 USA

Phone +1 520 881-0857 E-mail: [email protected] Fax: + 1 520 881-0554 http://www.radiocarbon.org/ 1999 PRICE LIST Proceedings of the 16th International Radiocarbon Conference $ 70.00* (Vol. 40, Nos. 1 and 2, 1998) INTCAL98 (1998 Calibration issue; Vol. 40, No. 3, 1998) 40.00 Proceedings of the 15th International Radiocarbon Conference (Vol. 37, No. 2, 1995) 50.00 Liquid Scintillation Spectrometry 1994 (ISBN: 0-9638314-3-7-,1996) 20.00 Liquid Scintillation Spectrometry 1992 (ISBN: 0-9638314-0-2; 1993) 10.00 (ISBN: 0-9638314-0-2; 1993) Special offer-LSC 92 and LSC 94 package-save $5.00 25.00 Late Quaternary Chronology and Paleoclimates of the Eastern Mediterranean 30.00 (ISBN: 0-9638314-1-0; 1994)

Tree Rings, Environment and Humanity (ISBN 0-9638314-2-9; 1996) 20.00 (Proceedings of the International Tree-Ring Conference, Tucson, Arizona, 1994) SUBSCRIPTION RATES VOLUME 41, Nos. 1-3, 1999 Institution 120.00 Individual 65.00 Lifetime Subscription-Institutional 2000.00 Lifetime Subscription-Individual 700.00 BACK ISSUES (except conference proceedings and special issues) Single issue 40.00 VOLUMES 1-9 each volume 40.00 VOLUMES 10-21 each volume 65.00 VOLUMES 22-38 each volume 100.00 Radiocarbon Conference Proceedings 50.00 SPECIAL FULL-SET OFFER-Volumes 1-40 (1959-1998) 800.00 Big savings. Includes 11 out-of-print issues. Take $50.00 off for each additional set.

POSTAGE AND HANDLING CHART

U.S. Foreign Postage rates are for surface mail. Please con- Subscription -- $10.00 tact us for airmail or express delivery rates. Book or Proceedings $2.75 $5.00 Orders must be prepaid. We accept payments Single back issue $1.25 $2.00 by Visa and MasterCard, or by check or money order payable in $US to Radiocarbon. Full set $35.00 $75.00

*Postage will be added; see above chart. Subscription rates and book prices are subject to change. The test of time

At the Rafter Radiocarbon Laboratory we have been successfully meeting the test of time for more than 45 years.

Athol Rafter established the laboratory in 1952. Today, using Accelerator Mass Spectrometry, we carry on the tradition of excellence that Athol Rafter began. At Rafter we understand what our clients expect - accurate dating, at competitive rates and superior turnaround times and service. Recent work undertaken by our team of multi-disciplinary scientists includes:

improving methods for contaminant removal in textile dating refining paleodietary studies improving techniques for pollen dating overcoming marine shell dating problems.

The Rafter Radiocarbon Laboratory has an international reputation for accurately dating a wide range of organic materials, sediments, textiles, bone, ivory, paper, wood, parchment, charcoal, shell, foraminifera and peat.

We also offer a wide range of archaeometric services that include stable isotope measurements (a13C, a15N, a180), amino acid profiles, PIXE/PIGME, X-ray diffraction, petrology and palynology. For accurate dating and analysis results that will stand the test of time, talk with Dr Rodger Sparks at the Rafter Radiocarbon Laboratory about your next project.

R ArR I C RADIOCARBON LABORATORY Institute of GEOLOGICAL & NUCLEAR Institute of Geological & Nuclear Sciences Limited SCIENCES PO Box 31312, Lower Hutt, New Zealand Limited Telephone: 64 4 570 4671, Facsimile: 64 4 570 4657

Email: r. sparks@gns. cri. nz cri. nz/atom/rafter/rafter. htm http://www.gns. IGN 7484