Ceramic Bristol

Direct Dating of Archaeological Pottery byCompound-Specific 14C Analysis of PreservedLipidsAndrew W. Stott,† Robert Berstan, and Richard P. Evershed*

Organic Geochemistry Unit, Biogeochemistry Research Centre, School of Chemistry, University of Bristol,Cantock’s Close, Bristol, BS8 1TS, U.K.

Christopher Bronk-Ramsey, Robert E. M. Hedges, and Martin J. Humm

Oxford Radiocarbon Accelerator Unit, Research Laboratory for Archaeology and the History of Art, Oxford University,6 Keble Road, Oxford, OX1 3QJ, U.K.

A methodology is described demonstrating the utility ofthe compound-specific 14C technique as a direct meansof dating archaeological pottery. The method uses auto-mated preparative capillary gas chromatography employ-ing wide-bore capillary columns to isolate individualcompounds from lipid extracts of archaeological pot-sherds in high purity (>95%) and amounts (>200 µg)sufficient for radiocarbon dating using accelerator massspectrometry (AMS). A protocol was developed and testedon n-alkanes and n-carboxylic acids possessing a broadrange of 14C ages. Analytical blanks and controls allowedbackground 14C measurements to be assessed and po-tential sources of errors to be detected, i.e., contaminationwith modern or dead 14C, isotopic fraction effects, etc. A“Russian doll” method was developed to transfer isolatedtarget compounds onto tin powder/capsules prior tocombustion and AMS analyses. The major advantage ofthe compound-specific technique is that 14C dates ob-tained for individual compounds can be directly linkedto the commodities processed in the vessels during theiruse, e.g., animal fats. The compound-specific 14C datingprotocol was validated on a suite of ancient pottery whosepredicted ages spanned a 5000-year date range. Initialresults indicate that meaningful correlations can beobtained between the predicted date of pottery and thatof the preserved lipids. These findings constitute animportant step forward to the direct dating of archaeologi-cal pottery.

Radiocarbon dating is an important analytical technique inarchaeology and is routinely used to date a wide variety of biogenicmaterials, e.g., wood, bone, and plant remains. However, many14C analyses of natural materials provide ages that only reflectthe bulk organic matter, which may be misleading in the case ofchemically complex heterogeneous samples. An example of thisin buried artifacts or ecofacts arises through contamination from

migration of older or younger carbon-containing organic matter,e.g., humic substances, thereby complicating 14C age-baseddeterminations. This problem of “sample/age homogeneity” canbe partially resolved through purification of the organic matterinto characteristic subfractions that directly equate to the originalsample, e.g., collagen from bone or cellulose from wood. However,contamination from the burial environment may still present aproblem even though rigorous preparation procedures have beenemployed. The ideal approach would be to date representativecompounds whose structures establish an unequivocal link to theirsource in antiquity. Previous attempts at molecular 14C dating haveincluded individual amino acids1 and peptides;2 however, theseare not regarded as routine analytical procedures. More recently,a compound-specific 14C dating approach has been demonstratedin the dating of sedimentary and petroleum-derived organicmatter.3,4 These researchers demonstrated the feasibility of usingpreparative capillary gas chromatography (PCGC) to isolate lipidsfrom complex mixtures in sufficient concentrations to allow 14Canalysis using high-precision accelerator mass spectrometry(AMS).

Relative dating of archaeological pottery has focused exten-sively on empirically derived sequences based upon changes inpottery characteristics with time, i.e., seriation. These chronologi-cal sequences form a routine part of establishing correlations anddifferences between areas and phases of archaeological sites.Absolute dates within these chronological sequences are oftenfixed using 14C dating of associated organic artifacts, e.g., boneor seed, or by dendrochronology. However, these chronologiesmay become vague if, for example, (i) unconformities arise in thearchaeological record, (ii) there is a lack of associated datableorganic matter, or (iii) misinterpretation of typologically basedceramic chronologies occurs.

* Corresponding author. E-mail: [email protected].† Present address: NERC 15N Stable Isotope Facility, CEH-Merlewood,

Grange-over-Sands, Cumbria, LA11 6JU.

(1) Bada, J. L.; Gillespie, R.; Gowlett, J. A. J.; Hedges, R. E. M. Nature 1984,312, 442.

(2) van Klinken H. J.; Hedges, R. E. M. Radiocarbon 1992, 34, 292.(3) Eglinton, T. I.; Aluwihare, L. I.; Bauer, J. E.; Druffel, E. R. M.; McNichol, A.

P. Anal. Chem. 1996, 68, 904.(4) Eglinton, T. I.; Benitez-Nelson, B. C.; Pearson, A.; McNichol, A. P.; McNichol,

A. P.; Bauer, J. E.; Druffel, E. R. M. Science 1997, 277, 796.

Anal. Chem. 2003, 75, 5037-5045

10.1021/ac020743y CCC: $25.00 © 2003 American Chemical Society Analytical Chemistry, Vol. 75, No. 19, October 1, 2003 5037Published on Web 09/03/2003

Potentially, one of the best methods of dating ceramics wouldbe to date organic matter directly associated with the pottery(either surface organic residues or organic matter absorbed withinthe fabric). Hedges et al.5 attempted to 14C date several organiccarbon-containing fractions from pottery with varying degrees ofsuccess. The variability in dates could be assigned to fractionimpurity and the presence of contaminants. The dating ofarchaeological ceramics thus presents a challenging goal withinthe archaeological discipline, specifically with regard to datingorganic matter directly associated with the pottery itself.

Organic residues in archaeological ceramics are the focus ofmuch ongoing research with the results obtained to date clearlydemonstrating that lipids can be consistently used as carriers ofimportant archaeological information relating to the following: (i)the nature of the commodities utilized in the vessels (Evershedet al. and references therein)6 and (ii) determining the mode ofvessel use.7 During this latter work, it was recognized that lipidsare preserved in ceramics in sufficient quantities to serve as asource of carbon for radiocarbon dating. Additional properties oflipids favor their use in 14C dating, namely: (i) their structures,distributions, and stable isotope values at natural abundancecan be confirmed by gas chromatography/mass spectrometry(GC/MS) and compound-specific stable isotope techniques, whichunambiguously define their origins and identify potential con-taminants, (ii) that lipids have fast metabolic turnovers and thusyoung ages at the time of deposition, and (iii) they are likely tobe largely indigenous to the ancient pottery vessels due to theirrelative immobility and hydrophobicity in the burial environment(soil). Up to now, radiocarbon measurements have been reportedfor two individual lipids isolated from an archaeological sample,namely, an oil from ancient Egypt. The compound-specific 14Cdates measured for the two compounds recovered from theEgyptian oil agreed reasonably well with the historic date of thefind spot from which the oil derived.3

This paper describes the results of a systematic study aimedat developing and testing an analytical protocol for the dating ofindividual lipids preserved in archaeological ceramics using acompound-specific 14C approach. n-Hexadecanoic (C16:0) andn-octadecanoic (C18:0) acids were targeted for analysis since theseare the most widely occurring lipids preserved in archaeologicalpottery. These compounds also persist in sufficient concentrations(10-4-10-3 g g-1 of potsherd) to make them realistic candidatesfor radiocarbon dating. PCGC was used to isolate individual fattyacids from lipid extracts of archaeological pottery in high purityand amounts sufficient for radiocarbon dating using AMS.

EXPERIMENTAL SECTIONTest Reagents and Samples. All archaeological samples are

listed in Table 1. High-purity solvents, glass distilled and HPLCgrade (Rathburns Chemicals Ltd.), were used during all experi-mental procedures and subsequent aliquots taken from the samebulk solvent. Authentic reference compounds, n-octadecane andn-heneicosanoic acids, were purchased from Sigma Chemicals Co.

Ltd. Glassware was either washed in Micro-90 (InternationalProducts Corp.) and rinsed with double-distilled water, acetone,and dichloromethane or heated in a furnace at 600 °C for 12 hprior to use.

Sample Preparation. Reference compounds, i.e., n-octade-cane and n-heneicosanoic acids, were dissolved in dichlo-romethane (1 mg mL-1 solution) prior to use. Fatty acids wereconverted to methyl esters using dry methanol (anhydrous Na2-SO4) acidified with concentrated H2SO4, heated at 70 °C for 1 h.Fatty acid methyl esters (FAMES) were then recovered by theaddition of 2 mL of double-distilled water and diethyl ether.Following removal of the diethyl ether and evaporation undernitrogen, the methyl esters were eluted through a short columnof anhydrous Na2SO4 for water removal, evaporated, and redis-solved in hexane.

Compositional analyses of lipids in archaeological potsherdslargely followed our previously published methodologies.7,8 How-ever, for the purposes of compound-specific 14C dating, Soxhletextraction [(dichloromethane/methanol (2:1 v/v)] of powderedsherds weighing >8 g was employed to recover larger amountsof total lipid extract (TLE) from which target compounds couldeventually be isolated. An aliquot of the TLE was derivatized withN,O-bis(trimethylsilyl)trifluoroacetamide containing 1% trichlo-rosilane (70 °C, 1 h) prior to screening by high-temperature GC(HT-GC) and HT-GC/MS. The remainder of the TLE washydrolyzed with methanolic NaOH (0.5 M, 70 °C, 1 h), acidifiedto pH 3 (HCl, 1 M), and extracted into hexane to yield a fattyacid fraction. An aliquot of this fraction was analyzed by GC-combustion-isotope ratio MS (GC-C-IRMS) to provide the δ13Cvalues of the individual FAMES before PCGC. Potsherd FAMESwere then submitted to PCGC as described below.

Instrumental Analyses. HT-GC, HT-GC/MS, and automatedGC-C-IRMS analyses were performed as described previously.8,9

Fully automated PCGC was carried out using a Hewlett-Packard7673 autosampler coupled to a Hewlett-Packard 5890 series II GCfitted with a Megabore fused-silica capillary column (30 m × 0.53mm i.d.) coated with a dimethyl polysiloxane stationary phase(DB-1, 0.5-µm film thickness). The GC was interfaced to a Gerstalpreparative fraction collector via a zero dead volume effluentsplitter connected to the flame ionization detector (FID) and atemperature regulated (200-300 °C) transfer capillary line inter-faced to the eight-port fraction collector. The zero dead volumepreparative fraction collector was installed in a temperature-controlled oven (250-300 °C) and consists of seven separatecapillary lines linked to individual glass U-tube traps (one waste,six sample traps). The autosampler, GC, and preparative fractioncollector are fully controlled by a microprocessor, which allowsprogrammable operation throughout the PCGC trapping se-quences. The GC temperature program was from 50 (2 min) to210 °C at a rate of 10 °C min-1 and then to 260 °C at 4 °C min-1

and finally to 300 °C at 10 °C min-1. Data were continuouslymonitored during PCGC run sequences via a Hewlett-Packard3396 series II integrator.

Samples were combusted at the Oxford Radiocarbon Accelera-tor Unit using a continuous-flow CHN analyzer (Europa-ANCA)(5) Hedges R. E. M.; Tiemei, C.; Housley R. A. Radiocarbon 1992, 34, 906.

(6) Evershed, R. P.; Dudd, S. N.; Charters, S.; Mottram, H.; Stott, A. W.; Raven,A.; van Bergen, P. F.; Bland, H. A. Philos. Trans. R. Soc. London B 1999,354, 19.

(7) Mottram, H. R.; Dudd, S. N.; Lawrence, G. J.; Stott, A. W.; Evershed, R. P.J. Chromatogr., A 1999, 833 (2), 209.

(8) Evershed, R. P.; Heron, C.; Goad, L. J. Analyst 1990, 115, 1339.(9) Copley, M. S.; Berstan, R.; Dudd, S. N.; Docherty, G.; Mukherjee, A. J.;

Straker, V.; Payne, S.; Evershed, R. P. Proc. Natl. Acad. Sci. U.S.A. 2003,100, 1524.

5038 Analytical Chemistry, Vol. 75, No. 19, October 1, 2003

Ta

ble

1.C

om

po

un

d-S

pe

cif

ic1

4C

Da

taM

ea

sure

do

nF

att

yA

cid

sE

xtr

ac

ted

fro

mM

ed

ieva

lto

Ne

oli

thic

Po

tsh

erd

s

sher

dsi

te/t

ype

targ

etfa

ttyac

ids

OxA

/OxA

-Xsa

mpl

eco

des

yiel

dC

O2

(µg

ofC

)co

mbu

sted

δ13C

FAM

E(‰

)δ13

CFA

(‰)

%m

od. ob

s

%m

od.

erro

r((

)

corr

ecte

d%

mod

.14

Ca

date

(er

rora

(yea

rsB

P)co

rrec

ted

date

a(2

σ)da

teby

asso

ciat

ionb

Wes

tCot

ton

Lyve

den

Aw

are

jar

RP2

2C

16:0

7355

29-

29.5

-28

.394

.01.

899

.820

(15

0m

oder

n12

50-

1300

AD

(S)

Wes

tCot

ton

St.N

eots

war

ebo

wlR

P73

C18

:075

8137

1-

29.6

-28

.683

.10.

587

.610

60(

5088

0-10

40A

D12

50-

1300

AD

(S)

Wes

tCot

ton

Lyve

den

AC

16:0

7586

27-

29.6

-28

.580

.90.

685

.912

30(

6067

0-96

0A

D12

25-

1250

AD

(S)

war

eja

rR

P25

C18

:075

8729

3-

31.7

-30

.886

.70.

691

.472

0(

6012

10-

1400

AD

C18

:175

8524

5-

27.8

-26

.786

.40.

791

.274

0(

6011

60-

1400

AD

Wes

tCot

ton

Shel

lyC

16:0

7582

79-

28.4

-27

.284

.90.

590

.085

0(

5010

40-

1280

AD

1150

-12

25A

D(S

)w

are

jar

RP7

8C

18:0

7583

416

-30

.5-

29.5

83.7

0.6

88.3

1000

(60

890-

1190

AD

C18

:175

8415

7-

28.6

-27

.586

.30.

791

.076

0(

6011

50-

1400

AD

Stan

wic

kM

orta

ria

C16

:075

9161

4-

28.6

-27

.478

.20.

583

.015

00(

5043

0-65

0A

D15

0-25

0A

D(S

)ST

-250

C18

:075

9248

9-

32.2

-31

.376

.70.

580

.917

00(

5021

0-44

0A

DSt

anw

ick

Mor

tari

aC

16:0

807-

0655

2-

29.2

-28

.079

.20.

584

.014

00(

5054

0-77

0A

D15

0-25

0A

D(S

)ST

-250

C18

:080

7-07

871

-32

.1-

31.2

77.5

0.4

81.7

1620

(40

260-

540

AD

Stan

wic

kM

orta

ria

C16

:080

7-08

558

-29

.1-

27.9

77.9

0.7

82.7

1530

(70

390-

660

AD

150-

250

AD

(S)

ST-2

50C

18:0

807-

0961

5-

32.2

-31

.376

.20.

580

.417

60(

5013

0-41

0A

DY

arnt

onM

ortla

kew

are

R31

C18

:080

6-07

326

-34

.2-

33.4

55.5

0.8

58.5

4300

(11

033

50-

2550

BC

3400

-29

00B

C(S

)

Yar

nton

Feng

ate

C16

:075

8959

9-

29.3

-28

.155

.70.

559

.142

20(

7030

20-

2570

BC

3400

-29

00B

C(S

)w

are

R5-

9C

18:0

7588

769

-30

.9-

30.0

56.3

0.5

59.4

4190

(70

2910

-25

70B

CC

18:1

7590

204

-30

.5-

29.5

56.9

0.8

60.0

4100

(11

029

50-

2350

BC

Eto

nR

owin

gLa

keC

16:0

7358

104

-31

.1-

30.0

57.1

0.8

60.6

4020

(11

029

00-

2200

BC

3700

-33

00B

C(S

)pl

ain

bow

lDB

C1

C18

:073

5624

2-

33.3

-32

.552

.60.

555

.547

30(

8036

50-

3360

BC

Eto

nR

owin

gLa

keca

rina

ted

bow

lD

BC

-E

C18

:084

9-08

280

-35

.9-

35.2

53.6

0.9

56.5

4580

(13

036

50-

2900

BC

4000

-36

00B

C(S

)

Ham

bled

onH

illpl

ain

bow

lHH

77-1

924

C18

:084

9-10

350

-32

.2-

31.3

53.9

0.6

56.8

4540

(80

3550

-29

00B

C48

70(

35B

P;37

10-

3540

BC

(A)

Ham

bled

onH

illpl

ain

C16

:080

6-10

442

-29

.228

.055

.30.

458

.742

80(

6030

90-

2670

BC

sam

ple

late

rth

anU

B-4

138

bow

lST

-81-

938

C18

:080

6-11

365

-33

.9-

33.1

53.8

0.6

56.7

4550

(90

3550

-29

00B

C(4

648

(21

BP)

,OxA

-704

2(S

tepl

eton

encl

osur

e)(4

735

(60

BP)

,ear

lier

than

OX

A-7

041

(448

5+/

-60

BP)

,OxA

-704

8(4

670

(40

(B

P)(A

)H

ambl

edon

Hill

C16

:080

7-12

511

-28

.8-

27.6

52.0

0.5

55.1

4790

(80

3700

-33

70B

Csa

mpl

ela

ter

than

UB

-413

8ca

rina

ted

bow

lST

-81-

96C

18:0

807-

1379

4-

31.1

-30

.251

.90.

454

.748

40(

6033

70-

3380

BC

(464

8(

21B

P),O

xA-7

042

(Ste

plet

onen

clos

ure)

(473

5(

60B

P),e

arlie

rth

anO

XA

-704

1(4

485

(60

BP)

,OxA

-704

8(4

670

(40

BP)

(A)

Swee

tTra

ckca

rina

ted

C16

:085

0-13

450

-29

.6-

28.5

51.9

0.4

55.1

4790

(60

3670

-33

70B

C38

07/6

BC

(D)

bow

lSW

2C

18:0

850-

1410

00-

34.4

-33

.651

.80.

454

.648

60(

6037

80-

3510

BC

aC

orre

cted

for

cont

ribu

tion

from

met

hano

l(de

riva

tizat

ion

reag

ent)

.b S,s

tylis

tic;A

,14C

date

ofas

soci

ated

mat

eria

l;D

,den

droc

hron

olog

y.

Analytical Chemistry, Vol. 75, No. 19, October 1, 2003 5039

fitted with a CO2 collection facility. The CHN analyzer uses GCto separate the CO2 from the other gases formed throughcombustion and collects the resultant CO2 as the target materialfor the gas source10 of the AMS system.

RESULTS AND DISCUSSIONThere were two main phases in the development of the

analytical protocol. The first phase focused on the implementationand testing of the analytical techniques required for the isolationof individual components from archaeological pottery. The secondphase, i.e., the AMS dating of individual lipids from archaeologicalpottery, depended upon whether acceptable controls and blankswere obtained. Validation of the method was achieved byradiocarbon dating individual lipids isolated from archaeologicalpottery of known date by correlation with materials from the samearchaeological contexts and related sequences dated by conven-tional methods.

Method Development. Analytical Blanks and Controls.Particular consideration was given to the suitability of the PCGCcapillary column used for the separation and isolation of individuallipids. The three key factors are (i) column dimensions, (ii)durability of the stationary phase, and (iii) reproducibility ofcompound retention times during repetitive trapping sequencestypically of >100 h. The GC was fitted with a Megabore fused-silica capillary column (30 m × 0.53 mm i.d.) coated with adimethyl polysiloxane stationary phase (DB-1, 0.5-µm film thick-ness) as the use of such columns had been previously shown toresult in minimal amounts of methylsilicone degradation productsaccumulating in the traps during a sequence.3 To assess thecontribution of carbon from the stationary phase, i.e., “columnbleed”, a chromatographic “blank” was trapped, i.e., collection ofthe entire chromatographic eluent from a sequence of 120 GCruns (temperature programmed to 300 °C), and submitted for 14Canalysis. Upon combustion, the amount of carbon (recovered asCO2) from the chromatographic “blank” was only 0.9 µg. As thechromatographic “blank” was trapped at temperatures well inexcess of the elution temperatures of the target compounds (<230°C) and each “blank” GC run was ∼40 times longer (34.5 min)than a typical target compound trapping window (∼0.8 min), itwas concluded that the amount of stationary phase (“columnbleed”)-derived carbon would be minimal in any compound-specific AMS analyses. This analysis also showed that contamina-tion from the instrument and carrier gas supply was negligible.Obtaining reproducible retention times for each compound isessential since in excess of 80 GC runs are normally required fora single trapping sequence. The drift in retention times duringseveral “>100 GC run” automated sequences was observed to varyby only 5 s. This confirmed the excellent reproducibility of theautomated instrument and thus enabled the precise setting of trapopening times during subsequent trapping sequences.

Our initial approach involved the analysis of appropriateanalytical controls in order to assess potential sources of con-tamination during manipulation and isolation of the individualcompounds. Four petroleum-derived solvents (dichloromethane,diethyl ether, hexane, methanol) were analyzed by AMS. Thesesolvents were an essential component of the analytical protocol,

being required for the extraction of lipids from the archaeologicalpottery and subsequent derivatization and transfer proceduresprior to AMS dating. The percentage of modern 14C in thesesolvents was <0.9%, and on this basis, all four solvents wereclassed as 14C “dead”.

Reference compounds including a C21 fatty acid (∼60% modern14C), modern cocoa butter, and a petroleum-derived n-C18 alkanewere chosen for development work because of their differing agesand GC properties similar to the compounds ultimately to beisolated from archaeological pottery. Analysis of these standardsenabled the assessment of whether extraneous 14C was beingintroduced. The quantity of modern carbon added in the methodwas less than 4 µg, with the amount of infinite age carbon addedless than 5 µg. The correction applied for date calculation was2.9 ( 0.3 µg of modern carbon based on the comparison of AMSmeasurements made on the bulk petroleum-derived n-C18 alkanewith measurements on smaller samples (down to 18 µg of C)combusted in the same way as the archaeological samples.

A concern at this stage was the potential isotopic fractionationoccurring during PCGC sequences. This was tested by determi-nation of the δ13C values of authentic compounds before and afterPCGC isolation. The n-C18 alkane was isolated after 100 PCGCruns (72-h run time) and its δ13C value measured. Comparison ofthe δ13C values before (-33.3‰) and after (-34.8‰) trappingsuggested a fractionation of 1.5‰, similar to that observed byEglinton and co-workers,3 i.e., <5‰. Similarly, the C21 fatty acidstandard showed a fractionation of 1.1% after 80 PCGC runs, i.e.-28.2‰ before and -29.3‰ after isolation. These data indicatethat fractionation does indeed occur to a small extent duringisolation, although the effect is minimized provided that (i) thePCGC instrument is carefully checked for leaks at the splitter,traps, and fraction collector and (ii) the trapping window issufficiently wide to span the whole elution time range of the targetcompound.

Trapping Recoveries and Sample Transfer to AMS. A splitratio of 98:2 was set in the effluent splitter using a separate controlcapillary installed below the jet of the FID. This directed 98% ofthe column effluent via a heated transfer line to the preparativefraction collector and 2% to the FID to enable continuousmonitoring during the PCGC sequence. Trapping recovery experi-ments were carried out using a stock solution of C16:0 and C18:0

fatty acids of known concentrations combined with a n-C18 alkaneinternal standard. By comparing the peak areas of the targetcomponents with the internal standard before and after trapping,the percentage recovery of each compound was calculated.Trapping efficiencies for the C16:0 and C18:0 fatty acids were foundto be >90%.

An important logistical component of the project was thetransfer of the isolated individual compounds from the PCGC trapsto the tin capsules for combustion to CO2. A method of depositingthe trapped target components onto tin capsules containing a smallquantity of tin powder was developed. The feasibility of the methodwas determined by injecting solutions of fatty acids of knownconcentrations (50, 100, 150, and 200 µg) onto tin capsulescontaining tin powder. By comparing the GC peak area responsesobtained from solvent washings with an internal standard, therecoveries of the fatty acids (C16:0 and C18:0) from the tin werecalculated to be >95%. Solvent creep (hence, sample loss) was

(10) Bronk-Ramsey, C.; Hedges, R. E. M. Nucl. Instrum. Methods B 1987, 29,45.


eliminated by employing a “Russian doll” method in which a tincapsule was partially filled with tin powder and then inserted insidea larger diameter tin capsule prior to spiking with the targetcompound. A key factor in the transfer method was the rapidsolvent evaporation (aided by heating at 40 °C) under a gentlestream of nitrogen during the spiking procedure.

Correction Factors and Dating. Additional corrections needto be applied to the Oxford AMS %modOBS value to account forthe effect on the measured radiocarbon age by the methyl carbonatom added through derivatization. These corrections are achievedas follows:

A simple mass balance calculation is used to correct for theinfinite age carbon added to the fatty acid from the methanol:

where no. of CFAME is the total number of carbon atoms inderivatized fatty acid, e.g., 19 for a C18 fatty acid, % modOBS is thepercent modern 14C of the derivatized fatty acid (corrected to δ13Cvalue, -25‰), and no. of CFA is the total number of carbon atomsin underivatized fatty acid, e.g., 18 for a C18 fatty acid.

To account for isotopic fractionation effects, each of theradiocarbon measurements is corrected to the accepted conven-tion δ13C value of -25‰. Since the AMS at Oxford measures the14C/13C ratio rather than conventional 14C/12C ratio measured byother laboratories, the isotopic fractionation effects on themeasured values are smaller by a factor of ∼2 (i.e., 1‰ variationequates to 8.2 radiocarbon years). The above isotopic fractionationcorrection is already incorporated into the % modOBS values;however, as the δ13C value of the original fatty acid is alteredduring the derivatization procedure, a further correction isnecessary. The δ13C value of the fatty acid (δ13CFA) is requiredand determined by a simple mass balance calculation:

where no. of CFAME is the total number of carbon atoms inderivatized molecule; δ13CFAME ) δ13C value of fatty acid methylester; δ13CMeOH ) δ13C value of methanol (-48.0‰); and no. ofCFA is the total number of carbon atoms in the original fatty acid.

By combining the correction factors due to the methanol (14Ccontent and δ13C alteration), an expression for the correctedpercent modern 14C (%modCORR) can be derived:

This is then expressed as a radiocarbon age (in years BP)according to the following equation:

A calibrated calendar range was obtained using OxCal v3.311

with the Stuiver et al. calibration curve.12

The previously described PCGC conditions were optimized toenable complete isolation of individual compounds in high purityand concentration. Trapped components were therefore guaran-teed to be free from any major 14C contamination, which potentiallymay have been introduced at any stage of the workup proceduresduring trapping and subsequent AMS analysis.

Validation of the Protocol Using Archaeological Pottery.The method was validated by dating individual lipids recoveredfrom archaeological pottery ranging in age from Early Neolithicto Medieval, a range spanning 5000 years. Sherds were madeavailable for analysis from a variety of archaeological sitesincluding the Raunds Area Project [West Cotton assemblage (LateSaxon/Early Medieval) and Stanwick (Roman)], Sweet Track(Early Neolithic), Eton Rowing Lake (Early Neolithic), HambledonHill (Early Neolithic), and Yarnton (Mid/Late Neolithic). SherdTLEs, derived from the processing or cooking of foodstuffs andcommodities, were initially screened by GC and quantified, andthose containing > 300 µg g-1 of lipid were selected as candidatesfor compound-specific 14C dating. The C16:0 and C18:0 fatty acids(mainly animal fat derived, based on the high relative abundanceof the latter component) were the most dominant components inall the selected TLEs. The δ13C values of these fatty acids weredetermined before and after trapping, from which it was evidentthat isotopic fractionation occurred to a small extent, with the δ13Cvalues always agreeing to within 2‰ and in the majority of caseswithin the analytical precision of the GC-C-IRMS instrument((0.3‰).

Medieval Vessels. Four samples (RP22, RP25, RP73, RP78)were selected from the West Cotton assemblage, Northampton-shire (Northampton Archaeology Unit) ranging in date from 1150to 1300 AD (dates assigned by typology and stylistic criteria;Blinkhorn, personal communication). Table 1 shows the dates ofeach of the sherds based on these criteria. Sherds from thisassemblage had previously been analyzed in our laboratory aspart of a larger study and were initially chosen because of theirhigh content of preserved lipids: RP25, 2031 µg g-1, RP73, 2821µg g-1, and RP78 4840 µg g-1. Sherd RP22 was chosen as theexception as it yielded only 200 µg g-1 of lipid and served as aguide in defining the lower limits of lipid yield that could besuccessfully used for PCGC. Determinations yielded targetcompounds ranging between 29 and 416 µg of combustible carbon(as CO2). The fatty acids from sherd RP22 were among the firstto be isolated and dated using the developed protocol. However,the low amounts of CO2 produced from combustion indicated that∼200 µg is the lower limit of sample size for valid compound-specific dating. Figure 1 shows the GC traces of trapped productscollected from 95 preparative GC runs from sherd RP25. Eachcompound has been isolated in the high purity essential forcompound-specific 14C analyses. Table 1 shows the 14C dataobtained for the most abundant fatty acids purified from the lipidextracts of the four Medieval archaeological potsherds by PCGC.

The very substantial pottery assemblage from West Cotton(>105 potsherds) has allowed typological chronologies to beascertained. A number of timber posts were also recovered from

(11) Bronk,-Ramsey, C. Oxcal v3.3, 1999.(12) Stuiver, M.; Reimer, P. J.; Bard, E.; Beck, J. W.; Burr, G. S.; Hughen, K. A.;

Kromer, B.; McCormac, G.; van der Plicht, J.; Spurk, M. Radiocarbon 1998,40 (3), 1041.

% mod(corrected for 14C content of MeOH) )no. of CFAME × % modOBS

no. of CFA

δ13CFA )(no. of CFAME × δ13CFAME) - δ13CMeOH

no. of CFA

% modCORR ) (no. of CFAME × % modOBS

no. of CFA )(1 + [(-25 + δ13CFAME)/1000]

1 + [(-25 + δ13CFA)/1000] )

radiocarbon years (BP) ) -8033 ln(% modCORR/100)


a sequence of water mills at the site and radiocarbon datedindependently. Sherd RP78 was recovered from a clay bankoverlying the infilled leat of the final watermill, which on the basisof the pottery would date the construction of the bank to aboutcal. 1150 AD (personal communication, A. Chapman, Northamp-tonshire Archaeology Unit). Timber posts (941 ( 53 BP; 88.9%mod. 14C, 1014 ( 51 BP; 88.1% mod.14C) from the survivingstructure of the mill yielded radiocarbon dates suggesting thatthe mill was in operation for about a century between 1025 and1125 AD. The Shelly Ware pottery associated with this construc-tion has been dated typologically to 1150-1225 AD. Compound-specific 14C dates on the C16:0, C18:0, and C18:1 fatty acids from RP78yielded dates of cal. 850 ( 50 BP; 90.0% mod. 14C (cal. 1040-1280 AD), 1000 ( 60 BP; 83.3% mod. 14C (cal. 890-1190 AD) and760 ( 60 BP; 91.0% mod. 14C (cal. 1150-1400 AD), respectively,which are in good agreement with the predicted age (Table 1).

Roman Vessels. Two fatty acids (C16:0 and C18:0) were isolatedfrom the total lipid extract of a Roman Mortaria (ST-250, datedtypologically at 150-250 AD) by PCGC. Upon combustion, yieldsof CO2 were 614 and 489 µg, respectively, with AMS analysisproducing dates of 1500 ( 50 BP; 83.0% mod. 14C (cal. 430-650AD) and 1700 ( 50 BP; 80.9% mod. 14C (cal. 210-440 AD) for theC16:0 and C18:0 fatty acids, respectively. The date for the C18:0 acidcorrelates well with the expected age of the vessel, i.e., 79.5% mod.14C. However, the C16:0 fatty acid yielded a slightly later date (>%mod. 14C). To investigate the reproducibility of this result, twofurther extractions, PCGC isolations and AMS analyses, wereperformed on the same sherd. The radiocarbon dates for theduplicate C16:0 fatty acids were comparable with those obtained

previously, i.e., later than second-third century Roman, and theC18:0 fatty acids again correlated well with the expected age ofthe vessel. The calibrated calendar date ranges for the fatty acidsare shown in the histograms displayed in Figure 2. The yields ofcombustible carbon for the replicate analyses exceeded 550 µgin all cases (Table 1); therefore, the difference in age cannot beexplained on the basis of sample size.

Neolithic Vessels. Three fatty acids (C16:0, C18:0, and C18:1)isolated in high purity by preparative GC (95 runs) from YarntonMid-Late Neolithic Fengate Ware. R5-9 yielded very consistentradiocarbon ages ranging between 59.1 and 60.0% mod. 14C (cal.3020 and 2350 BC); cf. Table 1. Two of the three compoundsyielded larger amounts of CO2 than had previously been recoveredfrom other vessels, i.e., 599 and 769 µg, which correlated wellwith the predicted amounts (∼550 and ∼685 µg) calculated froman external C21 fatty acid standard of known concentration runafter the PCGC sequence. The C18:0 fatty acid isolated from asecond sherd (R31, Mid-Late Neolithic Mortlake Ware) yielded aradiocarbon age of 4300 ( 110 BP; 58.5% mod. 14C (cal. 3350-2550 BC). All compound-specific 14C dates obtained for theYarnton pottery show that the individual lipids correlated withthe expected age of the vessels (Table 1).

Fatty acids (C16:0 and C18:0) were isolated in high purity froma plain bowl sherd (DBC-1) and a carinated bowl sherd (DBC-E)recovered from the Early Neolithic site at Eton Rowing Lake,Oxfordshire (Table 1), yielding, upon combustion 104, 242(DBC-1, C16:0, C18:0), and 280 µg (DBC-E, C18:0) of carbon,respectively. The compound-specific radiocarbon dates obtainedclosely correlate with the expected range of the Early NeolithicEton pottery, particularly those of the C18:0 fatty acids, i.e., 4730( 80 BP; 55.5% mod. 14C (cal. 3650-3360 BC) and 4580 ( 130BP; 56.5% mod. 14C (cal. 3650-2900 BC), respectively.

Early Neolithic pottery from the Hambledon Hill prehistoricmonument was sampled (Dorset County Museum) and screenedfor lipids. These sherds were chosen specifically because theyoriginated from contexts that contained bone or charcoal frag-ments that had been independently radiocarbon dated. Of the20 sherds initially screened, 40% yielded lipid, of which three(ST-81/938, ST-81/96, HH77-1924) were selected for compound-specific 14C analysis on account of their high lipid content.HH77-1924 was recovered together with other datable materials,which could be classed as “directly associated” with the sherd.ST-81/938 and ST-81/96 were recovered in secondary silts fromsegments 14 L3B and 15 L3B, respectively, if the sequence in othersegments were contemporary then the potsherds should be laterin date than the following artifacts: (a) UB-4138 (articulated dogskeleton, 4648 ( 21 BP), (b) UB-4151 (red deer antler pick, 4772( 19 BP), (c) UB-4152 (red deer antler pick, 4792 ( 20 BP), (d)UB-4153 (red deer antler rake, 4740 ( 19 BP), and (e) OxA-7042(antler, 4735 ( 60 BP). The potsherds should be earlier than thefollowing: (a) OxA-7048 (Maloideae charcoal, 4670 ( 40 BP), (b)OxA-7050 (Maloideae charcoal, 4730 ( 40 BP), and (c) OxA-7814(adult female humerus, 4680 (30 BP).

Compound-specific AMS dates measured on the target C16:0

and C18:0 acids resulted in values of 4790 ( 80 BP; 55.2% mod. 14C(cal. 3710-3370 BC) and 4840 ( 60 BP; 54.7% mod. 14C (cal. 3770-3380 BC) for sherd ST81/96 and 4280 ( 60 BP; 58.7% mod. 14C

Figure 1. Partial gas chromatograms of products collected from95 preparative GC runs in which the C16:0, C18:0, and C18:1 fatty acidswere targeted from a West Cotton Medieval sherd lipid extract. Theupper chromatogram corresponds to the total lipid extract while thosebelow are analyses of the contents of four traps after the sequencewas completed.


(cal. 3090-2670 BC) and 4550 ( 90 BP, 56.7% mod.14C (cal. 3550-2900 BC) for sherd ST-81/938. As mentioned previously, sherdHH77-1924 was found directly associated with OxA-8845 (reddeer antler crown, 4870 ( 35 BP), i.e., from the bottom of thesame ditch. Compound-specific AMS dates measured on the C18:0

acid resulted in a value of 4540 ( 80 BP; 56.8% mod. 14C (cal.3550-2900 BC). The overall compound-specific results on theEarly Neolithic pottery from Hambledon Hill correlated well withthe ages of the sherds; however, it is difficult to know the trueage of the contexts where the availability of associated preciselydatable materials is limited.

Lipids were extracted from an Early Neolithic carinated bowl(SW2) from the Somerset levels in south west England. Thepotsherd was discovered adjacent to an elevated wooden walkway,known as the Sweet Track, that spanned 2 km of the swamp land.This manmade structure has been accurately dated by dendro-chronolgy to 3807/6 BC and was believed to be only used for∼10 years.13 The C16:0 and C18:0 fatty acids were isolated in largeamounts from the sherd, producing yields of carbon uponcombustion in excess of 400 µg. Compound-specific radiocarbonages for the C16:0 and C18:0 fatty acids were 4790 ( 60 BP; 55.1%mod. 14C (cal. 3670-3370 BC) and 4860 ( 60 BP; 54.6% mod. 14C(cal. 3780-3510 BC), respectively. The calibrated calendar rangesfor both the isolated fatty acids, particularly the C18:0 fatty acid,compare very favorably with the associated dendrochronologicaldate for the Sweet track (Table 1 and Figure 3).

CONCLUSIONSThe experiments and determinations reported herein confirm

that PCGC is capable of isolating individual compounds from lipidextracts of potsherds via repeated injections, in sufficient amountsfor AMS dating (>200 µg of C). Through AMS analyses ofappropriate blanks and controls it has been established thatintroduction of background 14C contamination was minimal. AMSmeasurements on individual fatty acids isolated from potsherdsranging in date from Early Neolithic to Medieval yielded percentmodern 14C values, which were in good agreement with the samplehistory.

In general, the radiocarbon ages obtained for the C18:0 fattyacids correlated best with the associated ages, while a trendtoward a slightly more modern date was observed for several ofthe trapped C16:0 fatty acids. Interestingly, the triplicate measure-ments made on the Roman Mortaria vessel (ST-250) showed littlevariation, suggesting the age offset between the fatty acids doesnot derive from laboratory methodologies. Environmental con-tamination, involving migration of the C16:0 fatty acid (the mostdominant fatty acid in soil organic matter in this carbon numberrange) into the sherd during burial, could explain this apparentdiscrepancy. An enrichment of 14C has previously been observedin the bulk soil carboxylic acid fraction with increasing depthpresumed to result from the downward mobility of free fattyacids originating from modern vegetation, fauna, and microorgan-isms, during groundwater leaching.14 The somewhat enhancedmigration/leaching predicted for the more soluble shorter chainC16:0 fatty acid is consistent with the slightly younger valuesobserved for this component.

(13) Hillam, J.; Groves, C. M.; Brown, D. M.; Baillie, M. G. L.; Coles, J. M.; Coles,B. J. Antiquity 1990, 210.

(14) Bol, R.; Huang Y.; Meredith, J. A.; Eglinton, G.; Harkness, D. D.; Ineson, P.Eur. J. Soil Sci. 1996, 47, 215.

Figure 3. Radiocarbon age determination and calibrated calendarrange for the C18:0 fatty acid isolated from a carinated bowl sherd(Sweet Track, SW2). The calibrated calendar range is compared withthe associated dendrochronological date (3807/6 BC).

Figure 2. Calibrated calendar ranges from the analyses of the C16:0 and C18:0 fatty acids extracted from a Roman Mortaria sherd (ST-250)from the site of Stanwick. The C16:0 fatty acids (top three histograms) and the C18:0 fatty acids (bottom three histograms) were measured intriplicate and compared with the typologically expected date range for the sherd (lower panel).


The average age difference between the C16:0 and C18:0

components (in samples that produced >100 µg of C) was 216radiocarbon years. If the average age of the exogenous C16:0 fattyacid was half that of the endogenous component, the degree ofaddition to the potsherd would have to be as high as 5-15% ofthe exogenous soil C16:0 fatty acid. Considering the fatty acidconcentration difference between the potsherd and the soil is great(potsherd concentrations at least 10 times greater15), it is difficultto envisage that the addition of C16:0 fatty acid from the soil couldoccur to the required extent.

The complete data set presented above reveals disparitiesbetween some of the dates of the lipid from the potsherds and

those from the associated finds (Figure 4). When assessing theaccuracy of the compound-specific dates relative to the associateddatable artifacts, it is important to note that the reliability of thatassociation is dependent upon a number of variables affecting theformation of deposits at the archaeological sites, e.g., primaryversus secondary deposits, bioturbation, etc. As a result, cautionshould be exercised when comparing the compound-specific dateswith some of the associated finds. Possibly the best associateddate related to SW2, which was discovered adjacent to the SweetTrack. The compound-specific 14C dates for the isolated fatty acidsshowed excellent agreement with the age of the Sweet Track.Using 50 years as a more conservative estimate for the “uselifetime” of the Sweet Track and “decalibrating” its calendricaldate range into radiocarbon years, the midpoint of its range was

(15) Simpson, I. A.; van Gerben, V.; Perret, V. P.; Elhmmali, M. M.; Roberts, D.J.; Evershed, R. P. Holocene 1999, 9(2), 223.

Figure 4. Difference between the 14C ages of the individual fatty acids and the associated ages of each sample: (a) C16:0 fatty acids and (b)C18:0 fatty acids. Gray rectangles represent the associated age range, and the black bars correspond to the offset from the associated age.


found to be about 3 (160 14C years) and 5% (230 14C years) olderthan the date range midpoints that corresponded to the C18:0 andC16:0 fatty acids, respectively. These results are extremely encour-aging considering the Sweet Track was constructed over 5800years ago during the beginning of the Neolithic period of Britain.

As we proceed toward routine compound-specific datingstudies of individual lipids associated with potsherds it may bethat we will come to recommend that only the C18 componentneed be considered. Clearly, these findings constitute an importantstep forward in the direct dating of archaeological ceramics,targeting for the first time pure organic compounds derived fromorganic residues of commodities processed in the vessels duringtheir use.

ACKNOWLEDGMENTThe research was supported by a Natural Environment

Research Council (NERC) grant to R.P.E. and R.E.M.H. (GR3/10641). Jim Carter and Andy Gledhill are thanked for assistancewith GC/MS and GC-C-IRMS. NERC are also thanked for organic,stable isotope and accelerator MS facilities. We thank AndyChapman, Rob Perrin, Alistair Barclay, Tim Allen, Gill Hey,Frances Healey, Roger Mercer, Peter Woodward, and StephenMinnit for provision of pottery and advice on aspects of thearchaeology.

Received for review December 6, 2002. Accepted April 17,2003.

AC020743Y


Documents

Ceramic Bristol