Although the use of thermoluminescence (TL) for archaeological purposes was first mentioned as early as 1953 (by F. Daniels; cf. Michels 1973: 189), the initial practical uses of TL were in the detection of nuclear and radiation hazards (Fleming 1979). The term TL refers to the release of energy by crystalline solids when heated or exposed to light. Ionising environmental alpha, beta and gamma radiation results in the release of electrons and other charge carriers ('holes') in these materials. Electrons become trapped in defects of their crystal lattice, such as impurities or chemical substitutions. These metastable charge carriers accumulate over time at a known and largely constant rate determined by the dose of the radiation. They can be ejected from their 'traps' by an input of additional energy, causing them to recombine, which releases their excess energy as light, measurable in photons. This energy (TL) is therefore, with some qualifications, a function of the time since the material was last heated (e.g. ceramics or heating stones) or exposed to light (e.g. crystalline mineral grains).
TL dating made its debut in archaeology primarily to help in estimating the ages of pottery remains (Aitken et al. 1968, 1971; Fagg and Fleming 1970; Fleming 1968, 1971, 1979; Kennedy and Knopf 1960; Mejdahl 1969; Sampson et al. 1972; Zimmermann 1967, 1971). The use of its principles to determine when sand grains had last been exposed to sunlight is a more recent development (Aitken 1990; 1994; Huntley et al. 1985; Murray et al. 1997; Roberts and Jones 1994; Smith et al. 1990). It has been applied enthusiastically in Australia where archaeological dates exceeding 40 000 years were derived from two of the three luminescence methods now in use (Roberts and Jones 1994; Roberts et al. 1994, 1996). These are, besides standard TL analysis, optically stimulated luminescence (OSL) and infrared-stimulated luminescence (IRSL) analyses. The latter use either a green laser beam (OSL) or infrared light to free the trapped electrons.
Several technical difficulties apply to luminescence methods. An inherent problem concerns the annual environmental dose rate, which can be measured in the field or laboratory and which introduces the largest uncertainties into the method. For instance, substantial variability has been observed in K, Th and U, the principal sources of the environmental dose rate (Dunnell and Feathers 1994). Indeed, variations in sedimentary K may be directly related to former human occupation. For a field measurement the dosimeter would have to be placed virtually in the same location as the sample for a year, which is physically not feasible. Another difficulty concerns the moisture content, an important factor that cannot effectively be determined for the duration of the time in question.
Then there are specific problems relating to dating that relies on measurements taken from saprolithic or regolithic sediments, i.e. sediments that comprise grains from rock that decomposed in situ within the sediment. This is particularly crucial in sandstone shelters being formed by massexfoliation, one of the most common types of rock art sites. These shelters are usually not formed by fluvial, marine or tectonic action, but by the gradual exfoliation of degradation-susceptible sandstone facies. Rock fragments are prized from the walls and ceilings by Salzsprengung and other processes attributable to capillary moisture, they fall to the ground, become covered by sediment and then slowly decay due to ground moisture, becoming sand again. Only the surface grains of the rock fragments could have been exposed to daylight, while in the remainder of the material the 'TL-clock' has not been 'reset' (i.e. the traps have not been 'emptied' of electrons and holes). Therefore most of the sand formed by this process would show a TL age of millions of years, and any attempt to date such a sediment would combine essentially two groups of age readings. Significant errors through the misinterpretation attributable to this effect have already occurred in rock art dating, notably at the Australian site Jinmium. Here, archaeologists using TL analysis claimed an age of 58 000 to 75 000 years for petroglyphs that were clearly and obviously of the Holocene (Fullagar et al. 1996), and were subsequently shown to be so (Roberts et al. 1998). Such cases can readily be clarified by using OSL analysis instead, measuring each quartz grain separately and then discarding those results that are distinctly greater than the main cluster of data.
However, OSL dating, too, is not without significant qualifications. In TL dating it is traditionally customary to remove the outermost 2 mm of samples in the darkroom, to eliminate the need to account for dose rate alpha and beta radiation. This only penetrates to a depth in the order of microns, whereas gamma rays penetrate very much deeper. In the case of single-grain OSL analysis, this is obviously not possible, the grains are as a rule well under one millimetre in size. Their outermost rind may be removed by etching with hydrofluoric acid, but since the environmental radiation regime of the distant past cannot be known to us, absence of alpha-particle dose effects is not necessarily secured. Alpha and beta particles are far more ionising in their effects than gamma radiation. Consequently such results remain quite provisional, even in their order of magnitude, until they can be tested or the concerns can be dismissed. This applies especially to results secured from deposits on rock surfaces, such as mud-wasp nests. Disequilibrium in the uranium and thorium decay chains might occur more readily at such locations (Aitken 1985), and radon is very likely to be present in the sandstones concerned. These and other factors (changes in the hydrology due to past climatic changes) could have significant effects on past dose rates, which would yield correspondingly distorted age estimates. Examples of such experimental work are the OSL analyses currently being conducted on quartz grains from mud-wasp nests in northern Australia, some of which are said to be physically related to rock paintings (Roberts et al. 1997, 2000). Their study involves the added difficulties of having to assume that the mud-nests were built in adequate daylight to bleach all grains, or that the inner parts of the droplets of mud transported by the wasps were always sufficiently exposed to the sun. The first set of results reported consists of fifteen OSL dates ranging up to 1800 years BP, plus three dates of about ten times that age. It is hard to believe that such fragile structures should survive for up to 24 000 years, particularly as the authors mention no evidence of mineralisation. I have observed fully mineralised mud-wasp nests at Toro Muerto, a rock art site in central Bolivia, which are only of the late Holocene. Moreover, the claim that one anthropomorphous painting at the site is more than 17 000 years old is difficult to reconcile with Watchman's radiocarbon dates of the late Holocene, for similar figures in the same region (Watchman et al. 1997). Finally, it is rendered less credible by the fact that nowhere else in the world is there a Pleistocene painting tradition that has survived in such large numbers of motifs outside of caves.
It is to be hoped that these exciting claims from the Kimberley in Australia will withstand falsification attempts successfully, but for the time being the results will need to be considered carefully. They refer to the first attempt to introduce luminescence dating into rock art science and deserve every encouragement, but as with any pioneering endeavour of this type it is important that archaeologists exercise the requisite restraint in interpreting such preliminary and experimental results. Certainly this methodology is among the most promising in the age estimation of rock art. It applies not only to nests of mud-wasps, but also to similar structures by ants, termites and birds, all of which occur widely together with rock art.
A new development in rock art dating is the use of TL analysis to estimate the ages of calcite deposits in deep caves. Only two applications of this experimental method have so far been reported, one in Spain (Arias et al. 2000), the other in Piaui, Brazil. In the second case it was experimentally supplemented by electron spin resonance (ESR) dating, a method not so far used in rock art dating. If these approaches can be shown to provide accurate results it would be of considerable utility in determining the ages of cave art in both south-western Europe and Australia, where such art is most extensive and is frequently related physically to calcite speleothems.