Use of Luminescence Dating in Archaeology

INSTITUTE OF PHYSICS PUBLISHING MEASUREMENT SCIENCE AND TECHNOLOGY

Meas. Sci. Technol. 14 (2003) 14931509 PII: S0957-0233(03)56707-8

Use of luminescence dating in archaeologyJames K FeathersLuminescence Dating Laboratory, University of Washington, Box 353100, Seattle,WA 98195-3100, USA

E-mail: [email protected]

Received 27 November 2002, in final form and accepted for publication23 April 2003Published 29 July 2003Online at stacks.iop.org/MST/14/1493

AbstractWith improvements in methodology and instrumentation, luminescencedating is becoming a much more useful chronometric tool in archaeology.Procedures for dating ceramics are relatively routine and their accuracy hasbeen demonstrated in a number of studies. Research is aimed at applyingceramic dating in situations where other methods lack the direct datingability of luminescence to resolve chronological problems. Sediment datingis more difficult because of uncertainty in the extent of zeroing or because ofmixtures of different aged deposits. These problems have been addressed byisolating signals most likely to be zeroed and by dating single aliquots, andultimately single grains. Single-grain dating now has the potential not onlyto identify and date mixed deposits but to provide valuable information onsite formation processes. This is particularly critical for judging siteintegrity at controversial sites.

Keywords: luminescence dating, ceramics, sediments, archaeologicalmethod, single grains

1. Introduction

Until very recently archaeologists, particularly American ones,have looked upon luminescence dating with some scepticism(Feathers 2000a), which is somewhat ironic because it wasdeveloped in an archaeological context, in Europe in the1960s and 1970s, as a method for dating heated materials,primarily ancient ceramics. And although the dating contextlargely shifted to geological questions in the 1980s and 1990swhen application was extended to buried sediments, datingof sediments at archaeological sites nonetheless constituteda significant portion of this work. Still lingering doubtsabout accuracy and reliability have remained. In the last fewyears, however, the perception has begun to change. Morearchaeologists have been looking at it as a dating option,rivalling radiocarbon in some cases. This surge in popularitycan be measured in the United States by increased funding fromgrant sources such as the National Science Foundation, theinclusion of luminescence dating in archaeological contractsby government agencies and the proliferation of the numberof laboratories. (Ten at the last count in the United States,compared with only five a few years ago.) Similar expansionsare occurring elsewhere.

This sudden interest stems in no small part from bothmethodological advances, particularly in optically stimulated

luminescence (OSL), and improvements in instrumentation.Both have made luminescence dating more accurate and moreprecise. But it also must arise from an increased appreciationfor the intrinsic value for archaeology of luminescence dating,the ability to provide direct calendar dates for archaeologicalevents of interest. This paper reviews recent advancements ininstrumentation, method and application that have increasedthe value of luminescence dating for addressing archaeologicalquestions.

2. Background

Luminescence is the emission of light from crystallinematerials following the absorption of energy from an externalsource. It is distinguished from other light emissions suchas fluorescence by a time interval between absorption andemission, by a duration in a metastable energy state, thatpermits use for dating. The external source of energy cantake a variety of forms, but for dating purposes the sourceis naturally occurring, ionizing radioactivity (alpha, beta,gamma and cosmic radiation). Release of the absorbedenergy requires a stimulus, which again can have variedsources, but for materials used in dating the source is usuallyeither heat, resulting in thermally stimulated luminescence, or

0957-0233/03/091493+17$30.00 2003 IOP Publishing Ltd Printed in the UK 1493

J K Feathers

Figure 1. A simplified schematic of the physical process behind luminescence dating. At point A all traps (e.g. minerals in the raw materialof the pottery) have been filled. Between points A and B, a zeroing event (heating of pottery) empties the traps and in the processluminescence is emitted (unnoticed, of course, by the potter). Through exposure to natural radiation the traps are gradually filled throughtime, a process shown here as linear although it may not be at all points in time, particularly as saturation is neared. At point C to point D,the traps are again emptied, this time in the laboratory while luminescence is measured. The measured signal is proportional to the amountof trapped charge. If one knows how much radiation is required to arrive at point C, a value called the equivalent dose, then dividing by theaverage dose rate will yield the time from B to D.

more commonly thermoluminescence (TL), or light, resultingin OSL. The latter includes stimulation not only from thevisible region of the electromagnetic spectrum but frominfrared and ultraviolet radiation as well. Luminescence ismainly observed in ordered crystals, the metastable levelsassociated with crystalline defects that create localized chargedeficiencies. Quartz and feldspar are two minerals withsuitable luminescence properties commonly used in dating.

Luminescence is explained by reference to solid stateenergy band theory (Aitken 1985, 1998, McKeever and Chen1997). Ionizing radiation excites electrons from the valenceband, or ground state, across an energy gap to the conductionband. In ideal crystals energy levels within the gap cannotbe occupied, but defects allow metastable energy levels withinthe forbidden zone which can trap excited electrons or electronvacancies (holes). Traps are emptied of these charge carriersby stimulus from additional energy such as heat or light.The probability of escape, p, is governed by an Arrheniusrelationship, which for TL is expressed as

p = se(E/kT )

where E is the trap depth or the amount of energy requiredto move an electron from the trap to the conduction band (ora hole to the valence band), k is Boltzmanns constant andT is temperature. For OSL additional terms related to theintensity of light and the ability of the trap to absorb particularwavelengths are required. Upon escape the charge carrier mayrecombine with opposite charge residing at deeper traps calledrecombination centres and in the process release energy in theform of luminescence.

Materials with suitable luminescence properties can bedated because at some point in the past traps are emptiedof their charge by sufficient exposure to heat or light. Thisamounts to a zeroing event. If measurement of luminescencewas made immediately after this zeroing, no signal wouldbe observed. Subsequently, traps become refilled because ofcontinued ionization by radioactivity and a latent luminescence

signal steadily accumulates. When luminescence is measuredin the laboratory, this natural signal will be proportional insize to the time since the last zeroing event (figure 1). The ageequation is written as a simple ratio:

Age (ka) = DE (Gy)/DR (Gy/ka)

where DE is equivalent dose in grays (the unit for absorbeddose) and DR is the average dose rate over time. Time is givenas ka (1000 years) but other units of time could obviously besubstituted.

Equivalent dose, a concept sometimes difficult fornon-specialists to grasp, is essential for archaeologists tounderstand if they are to evaluate luminescence dates, sinceit is often the term with the greatest uncertainty. It is definedas the amount of radiation, in terms of absorbed dose, requiredto produce a luminescence signal equivalent to the natural onemeasured on the sample. In other words, how much radiationis needed to get from zero luminescence to the current,natural luminescence. Equivalent dose (DE ) is measured bycalibrating the natural signal against laboratory-administeredradiation, which acts as a proxy for natural radiation.

While the bulk of luminescence research is directed ataccurate estimates of DE , this is not to underestimate the equalimportance of an accurate estimate of the dose rate. The currentdose rate is often assumed to represent the average dose ratebecause of the long half-lives, of the order of 109 years, of themajor sources, 40K, 238U and 232Th. The dose rate includesan internal component (short-ranged alpha and beta radiation)from the sample itself and an external component (long-rangedgamma and cosmic radiation) from the environment. Theexternal dose rate can be difficult to estimate because ofcompositional inhomogeneity in the immediate environment(so-called lumpy environmentsSchwarcz (1994)). Alsoproblematic is a potential change in dose rate, either internalor external, through time either from a simple change in theamount of overburden or from a change due to other geologicalprocesses that add or subtract radionuclides. Such changes

1494

Use of luminescence dating in archaeology

can often be detected and sometimes corrected by observingstratigraphic trends (Mercier et al 1995a), by documenting thepresence of disequilibrium in the uranium and thorium decayseries (Readhead 1987, Olley et al 1996), or by comparingequivalent dose among different minerals or different grainsize fractions which experience differential dose rates (Vogelet al 1999, Feathers 2002).

TL is usually measured by heating a sample aliquot at asteady rate and detecting the light emission with a sensitivephotomultiplier. As the temperature increases, the probabilityof escape from traps increases while the number of electronsremaining in the traps decreases. This creates a peak for anyparticular kind of trap. The usual outcome is a compositeof peaks, relating to different traps, called a glow curve. Analternative method of measurement is to raise the temperaturerapidly to some point and hold it there, producing a decaycurve (called isothermal TL), which again is often a compositefrom several traps. OSL is usually measured as a decay curve,sometimes called a shine down, when the sample is illuminatedwith constant power.

Because traps have different energy depths, some traps arenot very stable at ambient temperatures and lose their electronsover the time period relevant to archaeology. Other trapsare sufficiently deep that they are not easily emptied duringthe depositional event. Obviously, neither very shallow norvery deep traps will be useful for dating, and an importanttask in dating is to isolate those traps which were emptied atthe time of interest and which have been stable since then.Since the natural signal is a product of only traps that arestable over the time of interest, and since signals inducedby artificial irradiation will include unstable components, thetask involves making these signals comparable, using variousthermal (preheat) treatments to simulate long periods of timeand plateau tests to gauge comparability.

3. Archaeological significance

Luminescence dates past exposure to heat and light, andbecause these exposure events are often the same eventsarchaeologists are interested in, luminescence dating has astrong advantage over other dating methods. Dean (1978)argued the importance, when interpreting dating results, ofdistinguishing dating events from target events. In tree-ringdating, for example, the dating event is the time when the treewas cut down, but the target event might be the use of thetree limb in construction. The two events might be widelyseparated in time, as, for example, in the cases of stockpiledwood or use of naturally fallen timbers, the so-called old woodeffect (Schiffer 1986). Bridging arguments are then neededto associate the dating event with the target event, argumentsoften required for interpreting dates from the most commonlyused chronometric method, radiocarbon dating.

In luminescence dating, the target events and the datingevents are often the same events, obviating the need forbridging arguments. In the case of ceramics, for example,the dating event is the last exposure to sufficient heat to zerothe signal. Firing temperatures of at least 500 C are normallyrequired, although temperatures as low as 300 C are sufficientif the duration of heating is long, and commonly the last suchexposure was when the ceramic was made or in some cases

when the ceramic was used, events that correspond preciselywith what archaeologists are interested in. The spectre of post-depositional heating, such as naturally occurring forest fires ormodern agricultural practices, is often raised but is rarely aproblem. Our laboratory has been involved in dating ceramicsfrom the North Carolina coastal plain, an area that has beensubject to frequent forest fires, but we nevertheless have beenable to achieve a reasonable chronology (Herbert et al 2002).We are currently dating ceramics from central Amazonia, anarea subject to modern slash and burn agriculture, to see howmuch of a problem that might create.

Heated lithics are somewhat different. While heattreatment was commonly employed in prehistory to improveflaking qualities, such heating was often not sufficient to resetthe signal. The manufacturing process may be directly datedif overheated lithics (wasters) can be located (Wilhelmsen andMiles 2003), but more often luminescence dating addresses adiscard event, when lithic tools became accidentally burned inhearths. Use of the hearths then becomes the dating event, anevent also of interest to archaeologists, and one that also canbe directly addressed by dating hearth stones, again providingthe fire was hot enough and the rocks thin enough to counterinsulating effects (Anthony et al 2001).

For sediments the dating event is the last exposure tosufficient sunlight. This is a depositional event. Unlike thecase of ceramics, however, where sufficient heat is nearlyalways applied in manufacture (there are rare exceptions forvery low-fired wares), the depositional agent of sediments maynot allow sufficient exposure to zero the signal fully. Thesignal may not be zeroed at all, so that a derived age wouldactually refer to some previous depositional event, or it maybe partially zeroed, retaining a sufficient residual from timeelapsed since the previous, full-zeroing event. For example,some colluvial depositional agents such as landslides may notexpose sediment grains at all to light. High-energy fluvialregimes, such as glacial outwash, because of rapid turbidityand high sediment loads may result in only partial exposureor differential exposure of grains (Gemmell 1997). Aeolianderived sediments are usually considered the best candidatesfor luminescence dating because airborne time allows amplesun exposure. Sufficient exposure, however, is governed bylocal circumstances not by any hard and fast rules aboutdepositional regimes. Our laboratory has had experience offluvial sediments having been well exposed and nearby aeolianones poorly exposed (Feathers 2003).

The fact that sediments are composites of uncementedgrains also makes them more problematic for dating thandiscrete objects like ceramics or lithics. A single deposit canresult from different transport processes, some of which maypermit wider exposure than others. Post-depositional mixingfrom turbation or deflation is also a major problem in datingsediments (Bateman et al 2003).

Despite these problems, which are becoming moreresolvable by methodological developments, sediment datinghas high potential for archaeological chronology. It provides adirect date for the deposition of artefact-bearing stratigraphiclayers, reducing reliance on the assumption that materials inthe deposits (such as charcoal) are the same age as the deposit.Luminescence may also allow dating of anthropogenicsediments. Artificial construction (walls, canals, mounds and

1495

J K Feathers

other earthworks) often has involved sedimentary material,humans being the transport agent (Stein 1987). If the materialwas exposed during construction, luminescence may provide ameans of directly dating such episodes (e.g. Bush and Feathers2003, Feathers 1997), or if the construction resulted in pits ortrenches, luminescence may address abandonment by datingthe immediate fill (Lang et al 1999, Sanderson et al 2003).

The discussions that follow are not a review ofluminescence work in archaeology but a selection of examplesthat illustrate particular applications. I have tried to selectexamples that postdate the exhaustive review of Roberts(1997). The examples are weighted to some degree towardsour work at the University of Washington laboratory, onlybecause of the authors familiarity. This literature can be wellsampled by reviewing issues of Radiation Measurements andQuaternary Science Reviews, two journals that publish manyluminescence applications, and by checking bibliographiespublished regularly in Ancient TL.

4. Dating ceramics

Luminescence dating first was developed in the context ofdating ceramics, using TL. The basic procedure was workedout by the late 1970s mainly by the Oxford laboratory underthe leadership of Martin Aitken (Aitken 1985, Feathers 2000a).Even after the development of OSL for dating sediments(Huntley et al 1985), TL has continued to be employed fordating ceramics. The main advantage of OSL in sedimentdating, the ease in isolating the most light-sensitive traps, is nota problem for ceramics. Still, OSL may have some advantagesfor ceramics in terms of requiring a smaller sample size, butlimited work has been done (Barnett 2000a, Guibert et al 2001,Hong et al 2001, Oke and Yurdatapan 2000), and often ceramicmaterials produce weak OSL signals compared with TL signals(Hong et al 2001) and may behave differently from unheatedmaterial (Barnett 2000a).

Luminescence dating commonly employs either polymin-eral fine grains or coarse silt- to sand-sized grains of eitherquartz or feldspars. Using coarse grains has the advantagesof eliminating the influence of short-ranged alpha irradia-tion, since neither quartz nor feldspar have appreciable inter-nal sources of alpha irradiation, and of involving only a sin-gle mineral with fairly well-understood luminescence proper-ties. Coarse-grain dating is still commonly used for ceramics(e.g. Barnett 2000a), but many ceramics lack abundant inclu-sions to make this practical. Our laboratory usually relies onfine grains. While having to include alpha irradiation in thedose rate and the complications that arise because of the lowerefficiency at trapping charge by alpha as opposed to beta andgamma radiation, fine grains can turn this to an advantage inthat less reliance is put on the more uncertain external doserate.

Two general techniques have been employed to measureDE by TL (figure 2). Both involve using several aliquots ofsample material, some of which are used to measure the naturalsignal and others of which are given calibrating irradiations.With the first, called additive dose, incremental irradiationsare given to different aliquots that still retain their naturaldose and a growth curve is constructed plotting irradiationagainst luminescence signal. The natural signal forms the

lowest point on this curve, which is then extrapolated backto zero dose to estimate DE . Such a procedure fails to takeinto account changes in the shape of the growth curve in theextrapolated region, and consequently additive dose, especiallyfor younger samples, frequently underestimates the age. Thesecond technique, called regeneration, applies incrementalirradiations to aliquots that have first been measured for theirnatural signal and thus zeroed. This procedure regeneratesthe growth curve from zero and the natural signal is fittedinto the curve by interpolation, the shape of the growthcurve not being as much of a problem. However, heatingof the sample to measure the natural signal often changes theluminescence sensitivity of the sample (i.e. it now takes more orless irradiation to produce the same amount of luminescence),so that the regeneration curve no longer mimics the originalgrowth curve of which the natural signal is a part. Interpolationthen leads to gross errors.

The traditional solution to the problems of both techniqueshas been to use the regeneration growth curve intercept tocorrect the extrapolated value from additive dose (the so-calledsupralinearity correction). Our laboratory has adopted themore precise slide technique, the idea for which was firstadvanced by Valladas and Gillot (1978) and Readhead (1988),and then developed for sediment dating by Prescott et al (1993)and Huntley et al (1993). Both additive dose and regenerationcurves are constructed. If the curves are the same shape (whichthey almost always are), then the regeneration curve can becorrected for sensitivity change by applying a multiplicativefactor to make the two curves parallel, a procedure that maynot be valid for samples close to saturation (almost never thecase for ceramics). Once this correction is made, the horizontaldistance (along the dose axis) between the curves can be takenas the DE . Our laboratory has dated over 500 ceramics in thismanner and has produced results that agree with independentevidence (where such is available) more than 85% of the time,perhaps higher if the independent evidence, whose accuracy isoften unknown, were truly accurate (e.g. Feathers 2000b).

Tests for accuracy involve evaluation of dates for eventsthat have been securely dated by independent means. Thisis not straightforward because seldom are the dating eventsprecisely the same for luminescence and the independentmethod, even if the target events are the same. An exampleworth detailing from our laboratory has been a comparisonof TL ceramic dates with tree-ring dates from several short-occupation Navajo sites in northwest New Mexico (Dykemanet al 2002, Feathers 2000b). From 12 sites both TL and tree-ring dates were available. Of the 15 TL dates, only 27% werewithin one standard deviation of the tree-ring dates, but a moretelling comparison was the aggregate data. Figure 3(a) showshistograms of the two sets of dates. Each TL date is treatedas a normal probability distribution and the intervals in thehistogram reflect summing of probabilities of any one datefalling in that interval. (The TL date distribution will havea broader distribution because of poorer precision than the1 year precision of tree-ring dating.) The distributions aresimilar with a main mode around AD 1700. An earlier minormode apparent in the tree-ring data is less defined in the TLdistribution. However, if the TL distribution is divided into thetwo ceramic types represented (figure 3(b)), the bimodality ofone of the types closely matches the bimodality in the tree-ring dates, while both types contribute to the main mode. The

1496


(a)

(c)

(b)

Figure 2. (a) Additive dose growth curve for sample UW736 (pottery from southern Arizona). Dose refers to artificial irradiation given inthe laboratory, so that the natural signal is located at the zero point. A linear fit to the points yields an extrapolated dose value of about0.8 Gy. (b) Regeneration growth curve for the same sample. The natural signal is represented by the triangle and interpolating that valueinto the linear fit of the points yields a dose value of about 1.15 Gy. (c) Both growth curves after they have been shifted horizontally intocoincidence with the best fit shown. The ratio of the two slopes is about 1.2 and this is used to correct for change in sensitivity. The amountof the shift (or slide) for the corrected points is about 1.0 0.08 Gy, taken as the best estimate for DE .

75th probability distribution of the TL dates for both ceramictypes (AD 16401780 for Gobernador Polychrome and AD15201740 for Dinetah Gray) is remarkably close to therange of production for both types estimated from radiocarbonand tree-ring dates from throughout the region (AD 16301775 for Gobernador Polychrome and AD 15401800 forDinetah Gray) (Dykeman et al 2002). This underscoresthe problem of comparing two techniques that date differentevents. The tree-ring dates address cut dates (at best) forthe trees ultimately used in construction, while the TL datesaddress ceramic production. The direct comparison showless correspondence because the events dated occurred atdifferent times, the distance between them being in part afunction of occupation duration. The aggregate comparisonsreflect regional activities over a block of time, where temporaldifferences in construction/production events may average out.Here the correspondence is much closer. This work alsodemonstrates that several dates (TL or otherwise) are necessaryto flesh out a chronology. One or two dates (of any kind)are not useful beyond a rough chronological indicator. Theaccuracy of this work is not an aberration but merely adds tothe growing evidence of the accuracy of TL dating appliedto ceramics (e.g. Barnett 2000a, Godfrey-Smith et al 1997,

Guibert et al 1994, Kojo 1991, Sampson et al 1972, Whittleand Arnaud 1975).

Precision is a separate problem. The lower resolvingpower of TL compared with radiocarbon or tree-ring datingis a function of the greater number of variables and intrinsicuncertainties involved (although a more precise techniqueis not necessarily a more accurate one, and TL retains itsadvantage as a direct dating method). Our laboratory routinelyproduces dates with better than 10% precision (Feathers2000b). These error terms reflect mainly analytical error,but also include systematic uncertainties on unmeasurablequantities such as past moisture contents (which affect theradiation dose the sample receives).

For many ceramic materials the fine-grain signal isdominated by contributions from feldspar, which have muchhigher luminescence sensitivity than other minerals such asquartz. A persistent problem in dating feldspars is anomalousfading, the non-thermal loss of stored charge from trapsthat from kinetic considerations should be stable at ambienttemperature (Spooner 1994a). Anomalous fading has beenattributed to quantum mechanical tunnelling (e.g. Visocekaset al 1994). By this theory a trapped electron has a smallbut finite probability of being found outside the energy barrier

1497

J K Feathers

(a)

(b)

Figure 3. (a) Comparison of tree-ring dates with luminescencedates. Each TL date is treated as a normal probability distributionand the intervals in the histogram reflect summing of probabilities ofany one date falling in that interval. (b) The same TL data, exceptdivided between two ceramic types.

that retains it. If this probability distribution overlaps thatof a recombination centre, the electron can escape, even atvery low temperatures. Such fading appears ubiquitous amongfeldspars, although some correlation has been found betweenthe amount of fading and the degree of disorder in the crystalstructure (Visocekas et al 1994). Feldspars of volcanic origin,where an ordered structure is prevented by rapid cooling,tend to have more problems with fading, which is one reasonceramics from areas with predominant volcanic geology (suchas Pacific islands) are difficult to date. Such regions also tendto have low radioactivity (Prescott et al 1982).

Fading can be detected by loss of signal as a functionof storage time and is routinely checked. Our experience isthat about 1 in 7 ceramics show evidence of serious fading.Quantum tunnelling predicts that the rate of fading will beproportional to 1/t where t is some unit of time, and thereforethe number of remaining trapped electrons decreases linearlywith ln(t). This is because centres and traps that are closetogether will fade first and as these are depleted the rate slows.This logarithmic relationship has allowed development of acorrection procedure that appears to work for samples thatare neither very young nor old enough that most fading has

been completed (Huntley and Lamothe 2001). The procedureuses fading rates measurable in the laboratory to estimatea percentage decrease through an interval of log time andthrough iteration calculates an age that with the estimatedfading rate would result in the derived, faded age. Thisprocedure has been tested on coarse feldspar grains usinginfrared stimulated luminescence (IRSL) in sediment dating(Huntley and Lamothe 2001), but our laboratory has tried usingit for fine-grained ceramics using TL. Although a systematicstudy remains to be done, initial results seem to indicate thatit works, although at a cost in precision (figure 4).

Other procedures have been proposed to deal withanomalous fading, mostly in the context of sediment datingbut probably applicable to heated materials as well. Lamotheand Auclair (2000) have found variable fading rates amongfeldspar grains from the same deposit and have used therelationship between equivalent dose and fading rate amonggrains to extrapolate to an equivalent dose with zero fading.Recent research in both TL and IRSL have identified a far-redto near-infrared emission band from feldspars that does notseem to exhibit fading (Zink and Visocekas 1997, Fattahi andStokes 2003). Isolating a red signal requires special attentionto photomultiplier and filter set-up to minimize red oven glow,in the case of TL, and interference from the stimulating light,in the case of IRSL. For ceramics, perhaps the simplest way tocircumvent fading is to extract quartz grains for dating. Sincefor many ceramics the amount of suitable quartz might besmall or difficult to isolate, further research into using OSLfor ceramics seems warranted.

Methodological research into ceramic dating has recentlybecome almost completely overshadowed by research intosediment dating (but see Barnett 2000b, Guibert et al 2001).Ceramic dating is generally considered routine, and mostlaboratories are oriented towards dating sediments. The mostpromising area of ceramic dating research in fact might beapplication, where advantage of the direct dating capabilitiesof luminescence dating can be taken. Unfortunately, little workdemonstrating this potential has been published, and the reviewby Roberts (1997) was able to list applications that mainlymade substantive rather than methodological contributions toarchaeology.

4.1. Mixed assemblages

A common problem in archaeology is isolating different agedcomponents from artefact concentrations that have been mixedby post-depositional processes (either natural or cultural) or byvirtue of being on an old surface. For example, pits of variousshapes and sizes are common features in many archaeologicalsites. Even where pits show no evidence of disturbance or slowinfilling, misleading conclusions can result from assuming thateverything in the pit is the same age. Yet radiocarbon dating ofcharcoal or other plant remains in pits is a common approachto attempt dating ceramics found in the same pits.

Barnett (2000b) has argued that the larger the pit the morelikely it contains materials of different agesin terms of theirorigin or manufactureeven if the infilling, or deposition, wasa single event. In smaller pits, the age of deposition is morelikely to correspond to the age of the contents. Using a seriesof TL dates on Iron Age/Roman sites in Great Britain she

1498


Figure 4. Examples of two samples showing evidence of anomalous fading. The table beneath the figure compares the corrected age, usingthe Huntley and Lamothe (2001) method, with expected ages derived from independent data.

showed that large enclosure pits contained ceramics datingto a wide range. Dating of the infilling could be estimated,short of dating the sediments themselves, by dating a largenumber of ceramics and using Bayesian statistics to estimatethe most recent date. In contrast, smaller pits from other sitescontained ceramics whose dates were much more constrained,especially after taking into consideration complications in theenvironmental dose rate due to differential radioactivity insideand outside the pit.

Holt and Feathers (2003) have shown that even smallpits can have problematic contents. The occurrence of twoceramic types, previously thought to be separated in time, inthe same pits at the BaehrGust site in the Lower Illinois RiverValley has puzzled local archaeologists, leading some to arguethat the ceramics were in fact contemporary. This impliedthat a regional trade and ceremonial network called Hopewellextended much later in the Lower Illinois than elsewhere.TL dating of pairs of ceramics from the same pit, however,clearly showed that one ceramic type was hundreds of yearsolder than the other and that the radiocarbon dates from somepits were highly misleading because they represented averagesof different aged charcoal. In this case, information on theenvironmental dose rate was not even available, but since theissue was the relative age between the two sets of ceramics,both of which experienced the same environmental dose rate,interval- rather than ratio-scale ages were obtained. By holdingseveral sources of possible systematic error constant (includingnot only the environmental dose rate but also moisture contentsand calibration of the laboratory irradiation source), intervaldating improved precision by as much as 25%.

Pits are not the only problem. Many sites in NorthAmerica, for example, occur on erosional surfaces, so thatthe entire record rests on the modern surface. Identifyingreoccupations is very difficult and radiocarbon is completelyuseless because the dated organics cannot be comfortablyassociated with any of the artefacts. Our laboratory isaddressing this issue at several sites in the Lower MississippiValley by dating a large number of ceramics from them. This

work is still in progress. An issue is the suitability of surfacefinds for dating because of uncertainties in the external doserate. Several studies have demonstrated the accuracy of datingsurface ceramics by TL (Dunnell and Feathers 1994, Sampsonet al 1997), suggesting the uncertainties from the external doserate are no greater than those from buried sites and perhaps lessso because of less uncertainty in the thickness and duration ofoverburden.

4.2. Site duration

Duration of occupation in any one locus is an importantvariable for understanding human demographic trendsand their impact on environment, economy and politicalcomplexity (Varien and Mills 1997). Controlling assemblageduration is also necessary when comparing similarities amongassemblages since variability will increase with duration.Estimating duration, however, has been a difficult problem,usually accomplished by an accumulation of radiocarbon ortree-ring dates. The problems of association are as acutehere as in any other dating endeavour (e.g. Nash 1997), andother archaeologists have attempted more direct measures suchas artefact accumulations, most parsimoniously expressed bythe Schiffer (1975) discard equation positing a relationshipbetween discard and occupation span (Varien and Mills 1997,Gallivan 2002). While gross comparisons are possible, lackof control over a number of variables often creates highuncertainties that cannot be quantified. Direct dating ofartefacts by luminescence provides a viable alternative. Datingseveral ceramic shards from one site will provide a directmeasure of duration of use that can be statistically quantified(e.g. by using the OxCal program of Bronk Ramsey (1995)).The more dates obtained, the higher the precision of siteduration. Precision can also be increased because onlyinterval-scale dating is required, as suggested earlier. Again,applications of this sort have not been published, but ourlaboratory is engaged in trying to determine the duration oflarge late prehistoric towns in the Mississippi Valley.

1499

J K Feathers

4.3. Small sites and other difficult-to-date assemblagesSince the 1960s, at least, archaeologists have been concernedwith securing a representative sample of the archaeologicalrecord. This has meant attention paid not just to large,conspicuous concentrations of artefacts but also to smallerconcentrations that dot the landscape. Where traditionalstudies relied on large-scale excavations, the representativeapproach has necessarily relied on surface survey. In mostplaces such survey has yielded numerous small artefactconcentrations. While important for understanding land usepatterns, these small sites have not attracted attention becausethey often lie completely within the ploughzone and aredifficult to date by traditional means. Small artefact samplesreduce the probability of recovering time diagnostic artefactsand organics, if preserved, cannot be associated with otherartefacts (Beck 1994). Without knowing how these sites fitinto a local cultural chronology, it is difficult to understandtheir significance. Our laboratory is beginning to address thisproblem by TL dating ceramics from small sites in severalareas in the Mississippi Valley, where late prehistoric largetowns are surrounded by numerous small sites, sometimescalled farmsteads. A presumption has been that the largesites represent nucleation of population from the earlier smallsites, but no one actually knows how old the small sites reallyare. Luminescence dates are providing the first indication.

Luminescence dating of ceramics has also proved useful inareas where other dating techniques are simply not applicable.Our laboratory has been dating ceramics from the sandhillsof coastal North Carolina, an area where almost no charcoalassociated with archaeological remains has been preserved andthe chronology has been poorly developed. Using TL, wehave worked out a tentative chronology that reasonably followstechnological trends and that is consistent with chronologicalwork in nearby areas (Herbert et al 2002). Barnett (2000a) haspublished TL dates of ceramics from Late Bronze Age and IronAge sites in England, a period of time when lack of stratigraphy,few diagnostic pottery styles and a relatively flat 14 C calibrationcurve have hampered development of a chronology. Her datesagree reasonably well with independent evidence. She alsonotes that chronologies based on technological properties ofthe pottery, what she calls fabric and includes non-plasticinclusions (temper), do not have much resolution. Whilearchaeologists have considered changes in temper during thisperiod in sequential terms, the TL dates suggest considerabletemporal overlap. Similarly, our laboratory has found thatthe supposed temporal sequence of limestone, grog and shelltemper in the southeastern United States is not an accurateportrayal. Limestone and grog tempers persist for a long periodafter shell temper is introduced (unpublished data).

5. Lithics

The direct dating advantage of luminescence dating ofceramics applies to lithics as well and methodologicalconsiderations are similar. The substantive contributions oflithic dating, however, are better known, so only brieflymentioned here, largely because the use of lithics extendswell beyond the chronological range of radiocarbon dating.In Palaeolithic studies, TL dating of lithics is often only

one of a few dating options available. TL has particularlybeen important in documenting the middle Palaeolithic andtransitions before and after (e.g. Mercier and Valladas 1994,Mercier et al 1995b, 1999, 2000, Valladas 1992, Valladas et al1998, 2001, Richter et al 2000, Porat et al 2002). Some lithicshave little internal sources of radiation, so dating requiresreliance on good estimates of the external dose rate, which mayinclude a high contribution from cosmic radiation (Richter et al2000). This will become particularly critical if lithic dating isextended to surface finds.

Of potentially wide applicability is using OSL or IRSLto date the light-exposed surfaces of rocks, lithic artefactsor stone architecture from archaeological sites. The datingevent is the last exposure to light of these materials, andcould refer to a burial event or, perhaps more interesting, tothe underside of lithic materials found on the surface. Thiswork is only beginning and will require special attention toboth collection procedures that prevent exposure (includingexcavating at night, Morganstein et al (2003)) and developmentof new measurement techniques and instrumentation to samplesurfaces (Habermann et al 2000, Greilich et al 2002, Theocariset al 1997).

This discussion on ceramics and lithics cannot be completewithout mentioning that luminescence may have additionalvalue for archaeology beyond dating. Luminescenceproperties have been used as sourcing criteria for determiningwhere material for artefacts originated (Akridge and Benoit2001, Hashimoto et al 2001) and to assess past thermaltreatments (Hashimoto et al 2001).

Mention also needs to be made of the application ofluminescence to other kinds of archaeological materials, suchas glass and metallurgical slag. Luminescence applications tothese materials have a long history (Aitken 1985), but recentstudies (e.g. Chiavari et al 2001, Gautier 2001, Haustein et al2001), while promising, suggest much work is still neededbefore such dating can be considered routine.

6. Sediments

Sediments are inherently more difficult to date than ceramicsbecause of the uncertainty of sufficient bleaching at depositionand because of the uncertainty in the integrity of the deposits.Heating unambiguously zeroes most ceramics, and the greaterchemical and erosional resistance of ceramics makes themless susceptible to post-depositional change. With sediments,only partial bleaching or no bleaching at all may occur atthe depositional time of interest and mixing of different-agedsediments may occur either during or after deposition. In eithercase, following a procedure similar to that for ceramics mayresult in gross inaccuracies.

Two general methods have been employed to dealwith partial bleaching or mixing. The first is to separateout components of the luminescence signal that are mostsusceptible to bleaching from sunlight and base dating onlyon them. This minimizes the problem of partial bleaching.The second is to date individually aliquots containing a smallamount of material, ultimately single grains. The distributionof ages among the aliquots or grains is used to identify mixing.

A number of studies (e.g. Bailey 2001, Huntley et al1996, Duller 1997, Krbetschek et al 1997) have shown that

1500


Figure 5. (a) TL regenerative glow curve of coarse-grained quartz from a sediment sample from a Brazilian rock shelter. The full curve isshown as a dashed curve, with the low-temperature peak reduced in scale. The slowly bleaching peak was obtained after an extended bleachwith green light. The rapidly bleaching peak is obtained by subtracting the slowly bleaching peak from the full curve (Feathers 1997).(b) OSL decay from another quartz sample. The inset shows three exponential components that when summed form the solid line throughthe data points in the main graph. The figure is modified from one presented by Bailey et al (1997).

the total luminescence signal from either quartz or feldspar isa composite of several signals arising from different traps orrecombination centres. Traps can be isolated by manipulationof the excitation source (heat or light), because of the differentenergy required to release the stored charge from differenttraps, while recombination centres can be isolated by selectionof emission wavelengths by optical filters.

OSL has gained popularity for dating sediments becauseit samples traps that are more susceptible to bleaching thanTL. In quartz, for example, the high-temperature TL signal ismade up of a rapidly bleaching peak and a slowly bleachingpeak (Franklin et al 1992, 1995) (figure 5(a)). The latter notonly is the result of slowly emptying traps but also containsan unbleachable residual. OSL is associated with only therapidly bleaching component (Spooner 1994b). While thetwo components can be isolated in TL (Franklin and Hornyak1990), it is much less time consuming to use OSL. Evenwithin OSL, however, there are components that bleach atdifferent rates (Bailey et al 1997) (figure 5(b)). Typically, OSLis stimulated using a narrow bandwidth of light of constant

intensity. Depletion of the trapped charge produces what iscalled a shine-down curve. If only one trap were involved,the curve would be exponential in shape, but fitting studieshave shown that the measured curves are a combination ofat least three, and maybe as many as five, exponentials, eachassociated with traps of different bleaching rates. To sampleonly the most rapidly bleaching component, which has thehighest likelihood of being emptied during deposition, datingstudies normally only utilize the initial small fraction of thecurve, and use the latter part of the curve as a subtractedbackground (Aitken and Xie 1992, Banerjee et al 2000).Comparing the DE derived from different parts of the shinecurve is one procedure used to detect partial bleaching (Baileyet al 2003). If the DE from latter parts of the curve, wheretraps more difficult to bleach are being sampled, match thosefrom earlier parts of the curve, an argument can be madefor adequate bleaching at deposition. More recently, somelaboratories have experimented with altering the intensityof stimulating light during the luminescence readout (Buluret al 2000). If the intensity is progressively increased from

1501

J K Feathers

Table 1. SAR protocol (Murray and Wintle 2000), taken from table 1 in Murray and Clemmensen (2001).1. Dose (= 0 Gy if natural signal)2. Preheat (e.g. 260 C for 10 s)3. OSL (e.g. 60 s at 125 C, stimulation using blue light, emission in UV), gives Li4. Fixed test dose (e.g. 5 Gy)5. Cut heat (e.g. 160 C)6. OSL (same as step 3), gives Ti7. Repeat steps 16 for a range of regeneration doses bracketing the natural dose, a zero dose and a repeat (or recycle) point8. Calculate Li/Ti for natural and regeneration points and construct regeneration curve to determine DE

zero to some maximum, called linear-modulated OSL, thedifferent components can be separated as peaks, making fittingprocedures mathematically more tractable (Agersnap Larsenet al 2000, Singarayer and Bailey 2003). For feldspars, ithas been found that the most bleachable traps are sensitive toinfrared radiation. Stimulation by such a long wavelength isexplained by a resonance effect (800900 nm) coupled with athermal assistance mechanism (Hutt et al 1988, Aitken 1998).IRSL has been widely used for dating feldspars, not onlybecause of rapid bleaching but because less attention is neededto separate feldspars physically because few other mineralshave this effect. (Despite problems with fading, feldspars havesome important advantages over quartzhigher sensitivity andhigher saturation levelsmaking them particularly suitable forolder samples.)

Perhaps more commonly encountered are deposits thathave been mixed with differently aged grains either duringor after deposition. To separate these different populationsof grains, only aliquots with a small amount of materialon them are dated. Numerous studies have shown that thepercentage of grains of quartz that are actually sensitive toluminescence (bright grains) is quite variable, between 5 and10%, although sometimes as high as 40% (Duller et al 2000,McFee and Tite 1998). Aliquots with a small amount ofgrains, say 1020 grains, are likely to have only one or twobright grains. If the deposit is made up of differently agedpopulations, this should be detectable from the DE distributionfrom several of these small aliquots, as evidenced by verybroad distributions or modality. A correlation of equivalentdose with brightness is also characteristic of differently agedor partially bleached samples, since brightness is a functionnot only of luminescence sensitivity but also of age (Clarkeet al 1999, Li 1994). The best resolution obtains when datingsingle grains, although requiring many more aliquots becauseso many grains have no signal. To make such measurementspractical advances in both method and instrumentation wererequired.

The method for determining DE outlined earlier forceramics requires multi-aliquots and is unsuitable for mixed-aged samples because of the averaging effects of using severalaliquots. A single aliquot additive dose process was proposedby Duller (1995). It involved measuring the natural signal andsuccessive signals from incremental doses, each preceded by apreheat, by using OSL shines that were short enough that littleof the total signal was removed at each step, thus allowingan additive dose build-up on the same aliquot. A correctionfor loss of signal at each step, especially from the successivepreheats, was necessary. The method suffered not only fromuncertainties in the correction procedure, especially for oldersamples that had begun to saturate and whose growth curves

were not linear, but also from the inherent problem of additivedosethe need to extrapolate to obtain DE and the subsequenthigh dependence on achieving the correct shape to the growthcurve.

Dullers early work was not successful in devising asingle aliquot regeneration method, because of the problemof sensitivity change. Once the sample had gone through onecycle of irradiation, preheat and measurement, its sensitivitychanged, because of either changed luminescence efficiencyper unit trapped charge or changed rate of trap filling perunit ionization, so subsequent signals from artificial doseswere not comparable. Murray and colleagues investigatedthis problem for quartz and through a series of experiments(Murray and Roberts 1998, Murray and Mejdahl 1999, Murrayand Wintle 2000) found that sensitivity changes could bemonitored by applying a small test dose after an OSLmeasurement. The signal from the test dose was found tohave the same sensitivity as the preceding OSL signal and socould be used as a correction to equalize the sensitivity of thenatural and subsequent regeneration signals. They deviseda procedure called single-aliquot regenerative-dose (SAR),which constructs a growth curve from regeneration signalswhose sensitivity is corrected to match that of the naturalsignal. The procedure was found to be relatively insensitiveto preheat temperature and size of the test dose. Because ofincreased accuracy and precision, SAR is now the preferredtechnique for most OSL applications involving quartz and isbeing extended to feldspar and polymineral fine grains as well(Wallinga et al 2000, Banerjee et al 2001, Auclair et al 2003,Fattahi and Stokes 2003).

The general sequence of steps in SAR for a single aliquot isgiven in table 1. The OSL measurement following the naturalor the regeneration dose is divided by the OSL measurementfollowing the test dose to equalize sensitivity. After severalregeneration cycles, the first regeneration dose is repeated tomake sure the correction procedure produces the same resultas the first time. A zero dose is usually included to assesswhether preheats or test doses are creating unwanted signals,called recuperation. It is also recommended to recover a knowndose by first zeroing an aliquot, giving it a dose, treating thatdose as if it were the natural signal and then using the procedureto see if the known dose can be recovered as the DE (Murrayand Wintle 2003). This ensures the procedure works for theparticular sample.

A variation of SAR was first applied to single grains ofquartz by Roberts et al (1998a, 1999) investigating a suspectedage overestimation from multi-aliquot analysis at Jinmiumrock shelter in Australia, which had placed human presencein Australia by 116 ka, far earlier than previously believed.By dating a large number of single grains, they were able

1502


to detect a complex distribution of ages, one part of whichsuggested an age of less than 10 ka (agreeing with a radiocarbonchronology), but also including a wide range of much olderages. The older ages were attributed to grains that had erodedoff the rock shelter wall and never been properly zeroed bysunlight. The younger ages were attributed to aeolian-derivedsediments that had been well zeroed and were believed torepresent the true depositional age.

The Jinmium work was done on conventional equipment,and although labour intensive, the study benefited from a largeproportion of bright grains. For many samples, the proportionof bright grains is much smaller and would have made thestudy impractical. Using aliquots with just a small number ofgrains, while increasing the likelihood of containing at leastone bright grain, loses some of the resolution provided bysingle grains. In 1999 Ris Laboratories introduced a specialattachment to their basic luminescence reader that provided ahigh-powered, finely focused laser that could stimulate singlegrains (Btter-Jensen et al 2002). The grains are arrangedon discs containing a grid of 100 holes, one for each grain.The discs can be irradiated and heated as a whole, while thelaser can rapidly measure the OSL from each grain. DaybreakNuclear, the other major supplier of luminescence equipment,is devising a similar version. This allows the SAR sequence tobe performed on a large number of grains in a relatively shorttime, depending on the irradiation lengths required. On someHolocene-aged sediments recently dated in our laboratory, anSAR sequence on 1000 grains could be completed in about24 h (Feathers 2003).

Dating single grains is not as straightforward as the abovediscussion suggests. Two problems make the interpretationof DE distributions obtained from large numbers of grainsdifficult. One is differential precision. Grains with thebrightest signals will produce values of DE with the greatestprecision (McCoy et al 2000). Even after eliminating grainswith no or very low signals barely above background, a rangeof ages will be obtained even from a single-aged populationbecause of this differential precision. Statistical displaymethods, such as histograms, which do not take this intoaccount, may be highly misleading. Galbraith et al (1999)introduced the use of radial graphs, which plot distributions asa function of precision, as a remedy (figure 6). The otherproblem is microdosimetry. The three major componentsof terrestrial radiation (gamma, beta and alpha rays) aredistinguished in part by their effective ranges. Alphas are shortranged but of little consequence for coarse-grained quartz orfeldspar. Gammas are long ranged and affect all grains within asample about equally. Betas, however, have a maximum rangeof about 2 mm. At the scale of single grains, the sources ofbeta irradiation are not likely to be evenly distributed, so thatsome grains will receive a higher beta dose rate than others.Measuring the differential rates at such a small scale is notcurrently possible, so the dose rate forming the denominator ofthe dating equation is only an average for all grains. Averagingwill underestimate the age for some grains and overestimate itfor others.

Errors arising from differential precision and micro-dosimetry should be random, so that a single-aged populationof single grains should have a distribution in DE that is nor-mal, with the mean representing the best estimate. This has

Figure 6. Radial graph for an aeolian sample from Mustang Springsin west Texas. Precision on the x-axis is the relative standard error.The y-axis is the number of standard deviations of individualequivalent dose values from single grains away from some referencevalue. The solid circles are those values within two standarddeviations of the reference. This is referred to here as the mean, butis not the mean of the distribution but rather that value thatmaximizes the number of points within two standard deviations. Acharacteristic of a radial graph is that a radial line drawn from theorigin will pass through points of equal equivalent dose but withdifferent precision. The radial line bisects the scale on the right atthat equivalent dose value.

been the working assumption, and the challenge is to deriveages from other distributions: multimodal, skewed or other-wise non-normal. For fluvial sediments, Olley et al (1999)hypothesized that because moving water entraps grains fromdifferent sources of different ages and can attenuate sunlight tosome degree so that only some of the grains are fully bleached,the result will be a sharp, positively skewed distribution with amode near the younger, or leading, edge. They proposed usingthe mean of the lowest 5% as the best estimate of DE and ob-tained reasonable results from some young fluvial sediments.Lepper and McKeever (2002) studied this in more detail andargued that the 5% was arbitrary and often underestimated thetrue age of the deposit. They proposed a leading edge statisticthat attempted to define a normal distribution that would char-acterize a lower mode. They did this by fitting a Gaussian curveto the leading edge of the distribution and taking the steepestpart of the slope (one standard deviation below the mode) asthe DE . The steeper the slope, the higher the precision. Othershave tried more sophisticated statistical approaches to isolatethe lowest, normal portion of the curve (Spencer et al 2003,Roberts et al 2000).

Post-depositional turbation that mixes sediments ofdifferent ages is a significant problem for sediment dating(Bateman et al 2003). The best estimate of DE from thetop of a stratigraphic unit, where older grains have beenturbated upwards, might be the leading edge, but the bestestimate from the bottom of the unit, where younger grainshave been turbated downwards, might be the trailing edge(Feathers 2003). Geological interpretation is clearly importantfor understanding these distributions, but conversely thesedistributions should also aid in geological interpretations,especially where independent age information is available.OSL dating of sediments could play a big role in understandingsite formation.

1503

J K Feathers

6.1. Anthropogenic sediments

Dating anthropogenic sediments is of obvious interest toarchaeologists. Direct dates for activities that produced thesediment obviate reliance on associated artefacts or organicmaterial. Burbidge et al (2001) used OSL on quartz to datethree soils created by post-Neolithic agricultural activities aswell as intervening sand deposits on the Shetland Islandsand achieved a stratigraphically consistent sequence consistentwith other archaeological evidence. Our laboratory has appliedOSL to soils that occur below and within artificial earthenmounds in order to date construction of these features (Bushand Feathers 2003, Feathers 1997). Dating of palaeosols byluminescence assumes that zeroing occurs by the recycling ofgrains to the surface by turbation processes. The mechanismfor exposure is probably erosion from microrelief created byprocesses such as tree-fall or animal burrowing. Once anactive soil is buried, surface exposure no longer occurs, so thatluminescence is dating the burial event, which for mounds isthe construction event. Presumably turbation is not successfulat bleaching all grains and the likelihood of exposure decreaseswith depth in the soil. We tested this assumption by lookingat modern top soils from the top of mounds and palaeosolsfrom within and under mounds and found positively skeweddistributions of equivalent dose, the leading edge, mean andrange of which increased with depth (figure 7). Mound fill,above and below the soils, produced older ages, suggestingthe mechanism of construction itself was not conducive tobleaching.

Other kinds of archaeological construction datable byluminescence are pits and canals, where dating of sedimentsat their base provides a minimum date of abandonment. Langet al (1999) used IRSL to date fine-grained sediments that hadwashed into prehistorically dug cellars from human-inducederosion. By determining DE from different parts of the shine-down curves, they found no evidence of partial bleaching.They also selected their samples to minimize the possibilityof admixture of older grains from the cellar walls. The derivedages were consistent with other archaeological evidence.

Sanderson et al (2003) have used both OSL and IRSL todate abandonment and use of ancient canals in the CambodianMekong Delta. Cursory luminescence measurements on aseries of small samples collected from a stratigraphic profilethrough one of the canals allowed separation of canal fillfrom the underlying substrate and also identification of laterdisturbance and other depositional features. More in-depthstudy of selected samples provided dates consistent withother archaeological data. Preliminary small-aliquot analysisshowed samples in the upper fill to be well bleached, butbasal samples gave evidence of substantial mixing with oldersubstrate (see also Spencer et al 2003). IRSL on feldspars fromthe basal samples also suggested that fine-grained feldsparsmay be better bleached in this depositional setting than coarsergrains or quartz.

Luminescence may also help in identifying anthropogenicsediments. Roberts et al (2001b) dated 19 fine-grained sam-ples from a thick loess deposit in northern China to determineaccumulation rates and to look for disturbance attributed toagricultural practices. The authors used blue stimulation fol-lowing a long IRSL exposure to minimize the OSL signal fromfeldspars and therefore the potential of fading. A modified

Figure 7. Histograms of equivalent dose distributions for segmentsof a modern soil at the Watson Brake archaeological site innortheastern Louisiana. The DE in the top 5 cm are clustered nearzero. Non-zero values become more common with depth as largerpercentages of grains have not been fully exposed through turbation.The data are from single aliquots containing 1090 grains, with theexception of the 510 cm segment where 200400 grain aliquotswere measured.

SAR protocol was adopted. The results showed that the depo-sition was quasi-continuous for at least 12 000 years. Fifteenof the dates were in the correct stratigraphic order. The fourexceptions, confined to two time periods, gave ages that weretoo high and were explained by agricultural activity that intro-duced unbleached grains into the deposits either by additionof material to increase arability or through human reworking.

6.2. Site integrity

Two archaeological issues that have engendered a gooddeal of debate are dates of colonization and correlation ofhuman activities with environmental events, such as theextinction of large mammalian fauna. Claims for earlycolonization or extinction correlations have been closelyscrutinized, particularly in terms of valid dating and siteintegrity. Luminescence provides not only an independentdating method but also, perhaps more importantly, anassessment of the amount of mixing or disturbance in thesediments. Interpretation of single-grain equivalent dosedistributions is already playing an important role in theAustralian debate. I have already mentioned Jinmium rockshelter, where single-grain distributions revealed a mixeddeposit that earlier was not appreciated and led to severe

1504


overestimation of the age of human occupation there. Incontrast, distributions from Malakunanja II rock shelter wereconsistent with a single bleaching date, indicating humanoccupation 5060 ka (Roberts et al 1998b). These ages wereconsistent with radiocarbon assays and agreed with earlier TLresults, which used multi-aliquot techniques and included asignal less sensitive to light.

The age of the oldest known human skeletons in Australia,from Lake Mungo, has been contested, with some OSL mea-surements, as well as U-series and ESR dating on the skeletonremains themselves, giving an age as old as 60 ka (Thorneet al 1999). Criticism of the OSL results has centred arounduncertainties in the environmental dose rate and the associ-ation of the dated sediments with the burials, but also overthe fact that the OSL was performed on multi-aliquots, whichcould mask incomplete bleaching (Gillespie and Roberts 2000,cf Grun et al 2000). Very recently a comprehensive OSL dat-ing project involving 25 dates from three stratigraphic sectionsbracketing two of the burials have constrained the dates of bothto 40 2 ka (Bowler et al 2003). Equivalent dose was deter-mined from both multi-aliquots, using the slide method, andfrom small (10 grains) single aliquots, using SAR. The meanresults were in agreement and the single-aliquot distributionsdid not indicate mixing or partial bleaching.

Roberts and colleagues have also begun an ambitiousproject to date the extinction of the Australian megafauna(Roberts et al 2001a). OSL samples were secured from 28palaeontological sites bearing on the issue and dated usingsingle aliquots containing typically less than 10 grains. Forsamples with positively skewed equivalent dose distributions,the youngest grains were taken to represent the equivalent dose,using a minimum age model (Galbraith et al 1999). They foundno evidence for the megafauna extending later than about 46ka and were therefore not able to rule out a human role inextinction. One potentially younger occurrence of megafaunaat the controversial site of Cuddie Springs was questionedbecause of evidence of sediment mixing, a conclusion stronglydisputed by the site excavators (Wroe and Field 2001).

OSL is also being applied to the contentious issue ofwhen people first settled in the Americas. Dating projectsare currently being carried out in Missouri, Virginia, SouthCarolina and Brazil at relevant sites, but so far little has beenpublished, except for TL dates on lithics and OSL dates onhearth sediments at an 11 ka site in the Amazon (Michab et al1998).

Finally, OSL is being increasingly used to evaluate the ageand integrity of sites associated with early evidence of modernhuman behaviour in sub-Saharan Africa (Henshilwood et al2002, Feathers and Migliorini 2001).

6.3. Site formationAn added benefit of sediment dating is information on siteformation processes, understanding of which is essentialfor not only assessing site integrity but to interpret spatialpatterning of artefacts.

Frederick et al (2002) used OSL dating to evaluate thedepositional process of an extensive sand mantle across eastTexas. The issue is important to archaeologists trying toassess the significance of archaeological material found in

Figure 8. Histogram of equivalent dose from an aeolian samplefrom Mustang Springs in west Texas, the same sample as illustratedin figure 6. The distribution is negatively skewed, suggestingextensive mixing in this deposit. The oldest grains, or trailing edge,yield ages closest to that expected from independent evidence(Feathers 2003).

these sands. One argument is that the sands are the product ofin situ weathering of Tertiary bedrock so that archaeologicalmaterials found at depth can only be there because ofpedoturbation, their spatial patterning therefore having noarchaeological significance. An opposing view is that thesands are aeolian Holocene deposits, which have preservedarchaeological spatial integrity. The authors reasoned that thesamples under the weathering model should have saturatedOSL signals but under the aeolian model should producesignals that were well bleached on deposition and that giveHolocene ages in correct stratigraphic order. Single-aliquotanalysis was used and while the aeolian model in generalwas supported, the distribution of equivalent doses did suggestsome post-depositional mixing on some samples, an issue theauthors are currently exploring (Bateman et al 2003).

Mixing from turbation or deflation in dune sedimentshas long been apparent to archaeologists and geologists(e.g. Mayer 2002). Figure 8 shows a negatively skeweddistribution of single-grain equivalent doses from a massiveaeolian deposit with little internal structure from the MustangSprings site in western Texas, suggesting extensive turbation.The trailing edge was used to date the sample. Althoughyounger than a single radiocarbon determination from the samelevel, this age is still preferable to the mean as an age fordeposition (Feathers 2003). The same study reported possibleturbation problems with other dune sediments, although otheraeolian depositions sampled, including some dunes, providedreasonable dates. Sommerville et al (2001) sampled a numberof dune deposits in Scotland, associated with archaeologicalsites, and found some variation in the extent of bleachingduring deposition, due in part to recycling from older dunedeposits, but nevertheless obtained ages that were broadly inagreement with independent evidence. Additional work onsmall aliquots (Spencer et al 2003) suggested there was somemixing. A statistical analysis of Gaussian distributions usingF-ratios applied sequentially to equivalent dose values (fromlow to high) suggested mixing from anthrogenic soils bothabove and below the aeolian deposit and may allow estimationof the time of abandonment between them.

1505

J K Feathers

Waterlain deposits from low-energy transport systemshave generally yielded successful results as well, but problemshave been reported. Folz et al (2001) reported OSLoverestimation when compared with radiocarbon dates from aLate Palaeolithic site located in Seine River deposits near Paris.While discounting several possible reasons to account for theoverestimation, they did not have the ability to date singlegrains. Possibly such resolution may have allowed detection ofthe positive skewness noticed in equivalent dose distributionsof other fluvial deposits (Olley et al 1999, Lepper et al 2000)and would imply a younger age.

Rock shelters have long been used for human occupationand in many areas because of preservation reasons provide alarge fraction of the archaeological record. Rock shelters serveas traps for sediment from a variety of sources, so the potentialfor complex radiation environments, mixing of sediments andpartial bleaching is high. Most OSL attempts have targetedaeolian-derived sediments in rock shelters (Schwarcz and Rink2001, Feathers and Bush 2000) but single-aliquot and single-grain dating should allow evaluation of colluvial sediments aswell.

7. Conclusion

Recent methodological developments in luminescence dating,particularly single-aliquot and single-grain dating, haveincreased accuracy and precision and have also extended theapplication of luminescence beyond dating to depositionalprocesses and site formation. Archaeologists are onlybeginning to take advantage of these developments and haveindeed been slow at even appreciating the benefits of traditionalTL dating of pottery that has been available for years.This paper has attempted to review these methodologicaldevelopments and to provide examples of applications thatdemonstrate how questions, important to archaeologicalmethod, can be addressed with luminescence. The directdating capabilities of pottery dating make luminescence anideal method for teasing apart mixed assemblages, determiningoccupation durations and dating small sites that are otherwisedifficult to place in regional chronologies. The resolution ofsingle-aliquot and single-grain dating presents the opportunityfor archaeologists to date and identify anthropogenic deposits,assess the stratigraphic integrity of sites and to determine siteformation processes. In many such applications, luminescenceshould become the dating method of choice.

References

Agersnap Larsen N, Bulur E, Btter-Jensen L and McKeeverS W S 2000 Use of the LM-OSL technique for thedetection of partial bleaching in quartz Radiat. Meas. 3241926

Aitken M J 1985 Thermoluminescence Dating (London: Academic)Aitken M J 1998 An Introduction to Optical Dating (Oxford:

Oxford University Press)Aitken M J and Xie J 1992 Optical dating using infrared diodes:

young samples Q. Sci. Rev. 11 14752Akridge D G and Benoit P H 2001 Luminescence properties of chert

and some archaeological applications J. Archaeol. Sci. 2814351

Anthony I M C, Sanderson D C W, Cook G T, Abernethy D andHousley R A 2001 Dating a burnt mound from Kilmartin,Argyll, Scotland Q. Sci. Rev. 20 9215

Auclair M, Lamothe M and Huot S 2003 Measurement of anomalousfading for feldspar IRSL using SAR Radiat. Meas. at press

Bailey R M 2001 Towards a general kinetic model for optically andthermally stimulated luminescence of quartz Radiat. Meas. 331745

Bailey R M, Singarayer J, Ward S and Stokes S 2003 Identificationof partial bleaching using DE as a function of illumination timeRadiat. Meas. at press

Bailey R M, Smith B W and Rhodes E J 1997 Partial bleaching andthe decay form characteristics of quartz OSL Radiat. Meas. 2712336

Banerjee D, Btter-Jensen L and Murray A S 2000 Retrospectivedosimetry: estimation of the dose to quartz using thesingle-aliquot regenerative-dose protocol Appl. Radiat. Isot. 5283144

Banerjee D, Murray A S, Btter-Jensen L and Lang A 2001Equivalent dose estimation using a single aliquot ofpolymineral fine grains Radiat. Meas. 33 7394

Barnett S M 2000a Luminescence dating pottery from laterprehistoric Britain Archaeometry 42 43157

Barnett S M 2000b Sampling pottery in luminescence dating studiesRadiat. Meas. 32 46772

Bateman M D, Singhvi A K and Frederick C D 2003 Investigationsinto the potential effects of pedoturbation on luminescencedating Q. Sci. Rev. 22 116976

Beck C 1994 Introduction Dating in Exposed and Surface Contextsed C Beck (Albuquerque: University of New Mexico Press)pp 313

Btter-Jensen L, Duller G A T and Murray A S 2002 Advances inluminescence measurements systems Radiat. Meas. 32 5238

Bowler J M, Johnson H, Olley J M, Prescott J R, Roberts R G,Shawcross W and Spooner N A 2003 New ages for humanoccupation and climatic change at Lake Mungo, AustraliaNature 421 83740

Bronk Ramsey C 1995 Radiocarbon calibration and analysis ofstratigraphy: the OxCal program Radiocarbon 37 425530

Bulur E, Btter-Jensen L and Murray A S 2000 Optically stimulatedluminescence from quartz measured using the linearmodulation technique Radiat. Meas. 32 40712

Burbidge C I, Batt C M, Barnett S M and Dockrill S J 2001 Thepotential for dating the Old Scatness site, Shetland, by opticallystimulated luminescence Archaeometry 43 58996

Bush D A and Feathers J K 2003 Application of OSL dating toanthrosols: a study of the distribution of DE of quartz throughsurface and buried soil profiles from Archaic-aged earthenmounds in the SE United States Q. Sci. Rev. 22 11539

Chiavari C, Martini M, Sibilia E and Vandini M 2001Thermoluminescence (TL) characterization and datingfeasibility of ancient glass mosaic Q. Sci. Rev. 20 96772

Clarke M L, Rendell H M and Wintle A G 1999 Quality assurancein luminescence dating Geomorphology 29 17385

Dean J S 1978 Independent dating in archaeological analysis Adv.Archaeol. Method Theory 1 22355

Duller G A T 1995 Luminescence dating using single aliquots:methods and applications Radiat. Meas. 24 21726

Duller G A T 1997 Behavioural studies of stimulated luminescencefrom feldspars Radiat. Meas. 27 66394

Duller G A T, Btter-Jensen L and Murray A S 2000 Optical datingof single sand-sized grains of quartz: sources of variabilityRadiat. Meas. 32 4537

Dunnell R C and Feathers J K 1994 Thermoluminescence dating ofsurficial archaeological material Dating in Exposed andSurface Contexts ed C Beck (Albuquerque: University of NewMexico Press) pp 11537

Dykeman D D, Towner R H and Feathers J K 2002 Correspondencein tree-ring and thermoluminescence dating: a protohistoricNavajo pilot study Am. Antiq. 67 14564

Fattahi M and Stokes S 2003 Absorbed dose evaluation in feldsparusing a single aliquot regenerative-dose (SAR) infraredstimulated red luminescence protocol Radiat. Meas. at press

Feathers J K 1997 Luminescence dating of early mounds innortheast Louisiana Q. Sci. Rev. 16 33340

1506


Feathers J K 2000a Luminescence dating and why it deserves widerapplication Its About Time: A History of ArchaeologicalDating in North America ed S Nash (Salt Lake City, UT:The University of Utah Press) pp 15266

Feathers J K 2000b Date list 7: luminescence dates on prehistoricand proto-historic pottery from the American SouthwestAncient TL 18 5161

Feathers J K 2002 Luminescence dating in less than idealconditions: case studies from Klasies River MainSite and Duinefontein, South Africa J. Archaeol. Sci. 2917794

Feathers J K 2003 Luminescence dating of sediments from theSouthern High Plains Q. Sci. Rev. 22 103542

Feathers J K and Bush D A 2000 Luminescence dating of MiddleStone Age deposits at Die Kelders J. Hum. Evolution 3891119

Feathers J K and Migliorini E 2001 Luminescence dating atKatandaa reassessment Q. Sci. Rev. 20 9616

Folz E, Bodu P, Bonte P, Joron J-L, Mercier N and Reyss J-L 2001OSL dating of fluvial quartz from Le Closeau: a LatePrehistoric site near Paris Q. Sci. Rev. 20 92733

Franklin A D and Hornyak W F 1990 Isolation of the rapidlybleaching peak in quartz TL glow curves Ancient TL10 16

Franklin A D, Hornyak W F and Dickerson W 1992 TL estimationof paleodose of dune-sand quartz Q. Sci. Rev. 11 758

Franklin A D, Prescott J R and Scholefield R B 1995 Themechanism of thermoluminescence in an Australiansedimentary quartz J. Lumin. 63 31726

Frederick C D, Bateman M D and Rogers R 2002 Evidence foreolian deposition in the sandy uplands of east Texas and theimplications for archaeological site integrity Geoarchaeology17 191217

Galbraith R F, Roberts R G, Laslett G M, Yoshida H and Olley J M1999 Optical dating of single and multiple grains of quartz fromJinmium rock shelter, northern Australia: Part I, experimentaldesign and statistical models Archaeometry 41 33964

Gallivan M D 2002 Measuring sedentariness and settlementpopulation: accumulations research in the Middle Atlanticregion Am. Antiq. 67 53553

Gautier A 2001 Luminescence dating of archeometallurgical slag:use of the SAR technique for determination of the burial doseQ. Sci. Rev. 20 97380

Gemmell A M D 1997 Fluctuations in the thermoluminescencesignal of suspended sediment in an alpine glacial meltwaterstream Q. Sci. Rev. 16 28190

Gillespie R and Roberts R G 2000 On the reliability of age estimatesfor human remains at Lake Mungo J. Hum. Evolution 3872732

Godfrey-Smith D I, Deal M and Kunelius I 1997Thermoluminescence dating of St Croix ceramics: chronologybuilding in southwestern Nova Scotia Geoarchaeology 1225173

Greilich S, Glasmacher U A and Wagner G A 2002 Spatiallyresolved detection of luminescence: a unique tool forarchaeochronometry Naturwissenschaften 89 3715

Grun R, Spooner N A, Thorne A, Mortimer G, Simpson J J,McCulloch M T, Taylor L and Curnoe D 2000 Age ofthe Lake Mungo 3 skeleton, reply to Bowler and Mageeand to Gillespie and Roberts J. Hum. Evolution 3873341

Guibert P, Schvoerer M, Etcheverry M P, Szepertyski B andNey C 1994 IXth millennium B C ceramics from Niger:detection of a U-series disequilibrium and TL dating Q. Sci.Rev. 13 55561

Guibert P, Vartanian E, Roque C, Schvoerer M and Bechtel F 2001Luminescence dating of burnt materials: effects of preheattreatment on OSL and consequences for dating proceduresRadiat. Meas. 33 43944

Habermann J, Schilles T, Kalchgruber R and Wagner G A 2000Steps towards surface dating using luminescence Radiat. Meas.32 84751

Hashimoto T, Komatsu Y, Hong D G and Uezu Y 2001Determination of radionuclides in small pieces ofarchaeological sample and its application to TL-dating Radiat.Meas. 33 95101

Haustein M, Krbetschek M R, Trautmann T, Roewer G andStolz W 2001 A luminescence study for datingarchaeometallurgical slags Q. Sci. Rev. 20 9815

Henshilwood C S, dErrico F, Yates R, Jacobs Z, Tribolo C,Duller G A T, Mercier N, Sealy J C, Valladas H, Watts I andWintle A G 2002 Emergence of modern human behaviour:Middle Stone age engravings from South Africa Science 295127880

Herbert J M, Feathers J K and Cordell A 2002 Building ceramicchronologies with thermoluminescence dating: a case studyfrom the Carolina Sandhills Southeastern Archaeol. 2192108

Holt J Z and Feathers J K 2003 Dating the Middle to Late Woodlandtransition in the Illinois valley: radiocarbon and TL dates fromthe BaehrGust site Midcontinental J. Archaeol. at press

Hong D G, Yi S B, Galloway R B, Tsuboi T and Hashimoto T 2001Optical dating of archaeological samples using a single aliquotof quartz stimulated by blue light J. Radioanal. Nucl. Chem.247 17984

Huntley D J, Godfrey-Smith D I and Thewalt M L W 1985 Opticaldating of sediments Nature 313 1057

Huntley D J, Hutton J T and Prescott J R 1993 The strandedbeach-dune sequence of southeast South Australia: a test ofluminescence dating, 0-800 ka Q. Sci. Rev. 12 120

Huntley D J and Lamothe M 2001 Ubiquity of anomalous fading inK-feldspars, and measurement and correction for it in opticaldating Can. J. Earth Sci. 38 1093106

Huntley D J, Short M A and Dunphy K 1996 Deep traps in quartzand their use for optical dating Can. J. Phys. 74 8191

Hutt G, Jaek I and Tchonka J 1988 Optical dating: K-feldsparsoptical response stimulation spectra Q. Sci. Rev. 7 3816

Kojo Y 1991 The reliability of thermoluminescence dating: a pilotexperiment Geoarchaeology 6 36774

Krbetschek M R, Gotze J, Dietrich A and Trautmann T 1997Spectral information from minerals relevant for luminescencedating Radiat. Meas. 27 695748

Lamothe M and Auclair M 2000 The fadia method: a new approachin luminescence dating using the analysis of single feldspargrains Radiat. Meas. 32 4338

Lang A, Kadereit A, Behrends R-H and Wagner G A 1999 Opticaldating of anthropogenic sediments at the archaeological site ofHerrenbrunnenbuckel, Bretten-Bauerbach (Germany)Archaeometry 41 397411

Lepper K, Agersnap Larsen N and McKeever S W S 2000Equivalent dose distribution analysis of Holocene eolian andfluvial quartz sands from central Oklahoma Radiat. Meas. 326038

Lepper K and McKeever S W S 2002 An objective methodology fordose distribution analysis Radiat. Prot. Dosim. 101 34952

Li S-H 1994 Optical dating: insufficiently bleached sedimentsRadiat. Meas. 23 5637

Mayer J H 2002 Evaluating natural site formation processes in eoliandune sands: a case study from the Krmpotich Folsom site,Killpecker Dunes, Wyoming J. Archaeol. Sci. 29 1199211

McCoy D G, Prescott J R and Nation R J 2000 Some aspects ofsingle-grain luminescence dating Radiat. Meas. 32 85964

McFee C J and Tite M S 1998 Luminescence dating ofsedimentsthe detection of high equivalent dose grains usingan imaging photon detector Archaeometry 40 15368

McKeever S W S and Chen R 1997 Luminescence models Radiat.Meas. 27 62561

Mercier N and Valladas H 1994 Thermoluminescence dates for thePaleolithic Levant Late Quaternary Chronology andPalaeoclimate of the Eastern Mediterranean ed O Bar-Yosefand R Kra (Tuscon: University of Arizona) pp 1320

Mercier N, Valladas H and Froget L 1999 Thermoluminescencedating of a middle Palaeolithic occupation at Sodmein Cave,Red Sea Mountains (Egypt) J. Archaeol. Sci. 26 133945

1507

J K Feathers

Mercier N, Valladas H, Froget L, Joron J-L and Ronen A 2000Datation par thermoluminescence de la base du gisementpaleolithique de Tabun (Mont Carmel, Israel) C. R. Acad. Sci.,Paris II 330 7318

Mercier N, Valladas H, Joron J-L, Schiegl S, Bar-Yosef O andWeiner S 1995a Thermoluminescence dating and the problemof geochemical evolution of sedimentsa case study: theMousterian levels at Hayonim (1994) Israel J. Chem. 3513741

Mercier N, Valladas H and Valladas G 1995b Flintthermoluminescence dates from the CFR Laboratory at Gif:contributions to the study of the chronology of the MiddlePalaeolithic Q. Sci. Rev. 14 35164

Michab M, Feathers J K, Joron J-L, Mercier N, Selos M, Valladas H,Valladas G, Reyss J-L and Roosevelt A C 1998 Luminescencedates for the paleoindian site of Pedra Pintada, Brazil Q. Sci.Rev. 17 10416

Morganstein M E, Luo S, Ku T-L and Feathers J K 2003 Uraniumseries weathering rind and luminescence dating of volcaniclithic artifacts Archaeometry at press

Murray A S and Clemmensen L B 2001 Luminescence dating ofHolocene aeolian sand movement, Thy, Denmark Q. Sci. Rev.20 7514

Murray A S and Mejdahl V 1999 Comparison of regenerative-dosesingle aliquot and multi-aliquot (SARA) protocols using heatedquartz from archaeological sites Quaternary Geochronology(Q. Sci. Rev.) 18 2239

Murray A S and Roberts R G 1998 Measurements of the equivalentdose in quartz using a regenerative-dose single-aliquot protocolRadiat. Meas. 29 50315

Murray A S and Wintle A G 2000 Luminescence dating of quartzusing an improved single-aliquot regenerative-dose protocolRadiat. Meas. 32 5773

Murray A S and Wintle A G 2003 The single-aliquotregenerative-dose protocol: potential for improvements inreliability Radiat. Meas. at press

Nash S E 1997 A cutting-date estimation technique for ponderosapine and Douglas fir wood specimens Am. Antiq. 62 26072

Oke G and Yurdatapan E 2000 Optically stimulated luminescencedating of pottery from Turkey Talanta 53 11519

Olley J, Caitcheon G and Roberts R G 1999 The origin of dosedistributions in fluvial sediments, and the prospect of datingsingle grains from fluvial deposits using optically stimulatedluminescence Radiat. Meas. 30 20717

Olley J M, Murray A S and Roberts R G 1996 The effects ofdisequilibria in the uranium and thorium decay chainson burial dose rates in fluvial sediments Q. Sci. Rev. 1575160

Porat N, Chazan M, Schwarcz H and Horwitz L K 2002 Timing ofthe lower to middle Paleolithic boundary: new dates from theLevant J. Hum. Evolution 43 10722

Prescott J R, Huntley D J and Hutton J T 1993 Estimation ofequivalent dose in thermoluminescence datingthe Australianslide method Ancient TL 11 15

Prescott J R, Robertson G B and Green J C 1982Thermoluminescence dating of Pacific Island pottery:successes and failures Archaeol. Oceania 17 1427

Readhead M L 1987 Thermoluminescence dose rate dataand dating equations for the case of disequilibrium in theuranium decay series Nucl. Tracks Radiat. Meas. 13 197207

Readhead M L 1988 Thermoluminescence dating of quartz inaeolian sediments from southeastern Australia Q. Sci. Rev. 725764

Richter D, Waiblinger J, Rink W J and Wagner G A 2000Thermoluminescence, electron spin resonance and 14C-datingof the late middle and early upper Palaeolithic site ofGeienklosterle Cave in southern Germany J. Archaeol. Sci. 277189

Roberts H M, Wintle A G, Maher B A and Hu M 2001b Holocenesediment-accumulation rates in the western Loess Plateau,China, and a 2500-year record of agricultural activity, revealedby OSL dating The Holocene 11 47783

Roberts R G 1997 Luminescence dating in archaeology: fromorigins to optical Radiat. Meas. 27 81992

Roberts R G, Bird M, Olley J, Galbraith R, Lawson E, Laslett G,Yoshida H, Jones R, Fullagar R, Jacobsen G and Hua Q 1998aOptical and radiocarbon dating at Jinmium rock shelter innorthern Australia Nature 393 35862

Roberts R G, Flannery T F, Ayliffe L K, Yoshida H, Olley J M,Prideaux G J, Laslett G M, Baynes A, Smith M A, Jones R andSmith B L 2001a New ages for the last Australian megafauna:continent-wide extinction about 46 000 years ago Science 292188892

Roberts R G, Galbraith R F, Olley J M, Yoshida H and Laslett G M1999 Optical dating of single and multiple grains of quartzfrom Jinmium rock shelter, northern Australia. Part II: resultsand implications Archaeometry 41 36595

Roberts R G, Galbraith R F, Yoshida H, Laslett G M and Olley J M2000 Distinguishing dose populations in sediment mixtures: atest of single-grain optical dating procedures using mixtures oflaboratory-dosed quartz Radiat. Meas. 32 45965

Roberts R G, Yoshida H, Galbraith R F, Laslett G M, Jones R andSmith M 1998b Single-aliquot and single-grain optical datingconfirm thermoluminescence age estimates at Malakunanja IIrock shelter in northern Australia Ancient TL 16 1924

Sampson C G, Bailiff I and Barnett S 1997 Thermoluminescencedates from Later Stone Age pottery on surface sites in theUpper Karoo S. Afr. Archaeol. Bull. 52 3842

Sampson E H, Fleming S J and Bray W 1972 Thermoluminescentdating of Colombian pottery in the Yotoco style Archaeometry14 1218

Sanderson D C W, Bishop P and Stark M 2003 Luminescencedating of anthropogenically re-set canal sediments fromAngkor Borei, Mekong Delta, Cambodia Q. Sci. Rev.22 111121

Schiffer M B 1975 Archaeology as behavioural science Am.Anthropol. 77 83648

Schiffer M B 1986 Radiocarbon dating and the old wood problem:th

Use of Luminescence Dating in Archaeology

Documents

Transcript of Use of Luminescence Dating in Archaeology