Figure 6. Radiocarbon ages for 135 of the 137 samples obtained from beeswax figures in northern Australia, after Nelson et al. (2000).

The state of our knowledge may seem somewhat better for pictograms than for petroglyphs, but impressions can be deceptive. The term pictogram is used here in accordance with the IFRAO Rock Art Glossary (Bednarik 2000c): a rock art motif that involved an additive process in its production, such as the application of paint, dry pigment or beeswax; it includes essentially paintings, stencils, drawings and imprints. The basic division of rock art into petroglyphs and pictograms is useful not only in technological considerations, but analytical methods are also frequently related to it. Attempts to recover the mineral dust produced in petroglyph production have remained unsuccessful so far (despite attempts in Australia by George Susino), hence there is generally no substance available for dating, whereas such substances generally do exist in pictograms. Moreover, preservation and site management techniques may also differ between these two divisions.

While the availability of datable substances relating to the making of pictograms may foster an expectation that more reliable results can be secured from this form of rock art, this is not necessarily reflected in the results tendered so far. They range from very reliable to unconvincing and probably false claims, so it is important to consider each result on its own merits. This applies especially to startling or sensational claims, which differ significantly from expectations, as has been the case in several projects of recent years.

Beeswax figures are a local but widespread feature in northern Australian rock art, where they have been reported from western Arnhem Land (Brandl 1968; Chaloupka 1993), Kimberley (Welch 1995), Reynolds River, Keep River and Groote Eylandt. Of all the rock art types in the world, these figures are perhaps the most amenable for dating. Not only is there always adequate datable substance available, the physical properties of the beeswax render it unlikely that it would have been used in any but nearly fresh condition, so the time of the production of the wax, which approximates its radiocarbon age, is unlikely to differ significantly from the time the art was created. Numerous such dates have already been secured from beeswax figures (Nelson et al. 1995; Tacon and Garde 2000). Nelson (2000) reports radiocarbon age measurements from 137 figures from sixteen sites in the Northern Territory. They range from the present time to about 4000 BP, but the overwhelming majority of the ages falls within the last 500 years, and all except four are under 1500 years. The apparent rarity of the few significantly older samples as well as the dates’ chronological distribution suggest that this is a result of taphonomic processes. The large number of available results renders it possible to detect a typical temporal distribution curve implying a taphonomic threshold (Bednarik 1994a) of about 800 years BP for beeswax figures (Figure 6, above).

Residues of rock paintings, stencils or drawings have often been concealed by accretionary deposits, such as carbonate, silica and oxalate skins or crusts. These and the inclusions they frequently contain have been used to secure data intended to assist in estimating pictogram ages. In interpreting such data it is important to appreciate the many qualifications that apply to them (see also comments elsewhere on this site). The most spectacular results of such work are from Australia, where at several sites series of superimposed painting events, separated by deposition of accretions, have been analysed, and Watchman (2000) has even detected paint residues in substrates that lack any surface indication of the presence of rock art. At Walkunder Arch Cave, Queensland, a series of ten radiocarbon dates has been obtained from finely stratified accretions totalling only 2.11 mm thickness (Campbell 2000). These dates are in sequence and range from 29 700 to 3340 carbon-years, spanning a history of about 26 000 years. The sequence includes three painting events, at 28 100 +/- 400, 16 100 +/- 130 and 10 400 +/- 90 carbon-years BP. The extraordinary precision of these nano-stratigraphic techniques and the availability of sequences rather than individual ‘dates’ certainly enhances the credibility of such results, but they are still subject to a variety of qualifications. Most important of all, until radiocarbon sampling can be focused on specific compounds or substances, i.e. at the object or molecular level (Bednarik 1996), we cannot know what it is we are dating, and the ubiquitous presence of datable carbon in rock substrates, together with the openness of the carbon system limits reliability of this approach.

The issue is well illustrated by another Australian project. Loy et al. (1990) claim to have identified blood residues at two Australian rock art sites. However, there are serious problems at the site Laurie Creek (Northern Territory). Underlying sub-modern paintings, ‘fragmentary panels of weathered dark red pigment’ were reported from which Loy secured a proteinaceous substance he identified as human blood. An accelerator mass spectrometry radiocarbon date of 20 320 + 3100/ -2300 years BP was obtained from this substance. His co-author Erle Nelson, however, had ‘second thoughts’ about the results and returned to the site to re-sample the surface deposit (Nelson 1993). He found that the reported pigment layer was nothing more than naturally reprecipitated iron oxides of a type common on sandstone surfaces, and he detected organic matter at various surface locations that bore no paint. When he re-analysed the deposit from which the original data had been obtained, he found only very low concentrations of protein. He (1993: 894) reports that ‘the material dated was not proteinaceous, and therefore not a remnant of human blood. … It is not a date with any archaeological meaning’. Loy (1994) has respon-ded to Nelson’s correction, still maintaining that there was mammalian IgG present at the sampling site, and saying that Nelson’s new data confirm the presence of organic carbon in the samples. Indeed they do, but the organic carbon is also present in the supposedly undecorated rock surface. However, Loy’s views have been firmly refuted by Gillespie’s (1997) subsequent research (cf. also Tuross and Barnes 1996).

Amino acid racemisation may have valid applications in rock art dating, despite the severe limitations imposed on this method by the extreme susceptibility of the reaction to temperature (Murray-Wallace 1993), and despite Denninger’s (1971) refuted attempt to date South African rock paintings. Amino acid residues can certainly be preserved in rock paints (McCarthy et al. 1994).

Attempts to secure ‘direct’ radiocarbon dates from paint residues are based on the assumption that their organic content reflects binders or incidental vegetable inclusions that are of essentially the same age as the rock art. It was first introduced in South Africa (Van der Merwe et al. 1987; Hedges et al. 1987). Binders were frequently used in paint preparations, and presumed brush fibres and other material relating to paint preparation have been detected (Cole and Watchman 1992; Watchman and Cole 1993). However, unless the substances concerned are isolated and identified before processing for carbon analysis, there is no certainty that such radiocarbon dates refer to the age of the paintings concerned (Bednarik 1996). The most commonly used rock art pigments are ochres, i.e. oxides or hydroxides of iron, and Ridges et al. (2000) have shown that samples of such minerals from rock art regions comprise substantial amounts of organic matter, such as lichen and bacteria. Ridges et al. detected phototropic microbes which could have radiocarbon ages that are unrelated to their actual antiquity, through recycling old organic carbon. Of particular concern should be their detection in one sample of approximately 5000-year-old unidentified organic matter. Had this sample been used in a rock painting, it would have provided a severely misleading carbon age. On the other hand, applied paints can be the target of bacteria and algal growths, as we know only too well from some decorated Palaeolithic caves in south-western Europe (but many other cases are known). Moreover, the ubiquitous presence of organic matter in probably most lithospheric substrates (Bednarik 1979; Nelson 1993) questions the reliability of such dates in any case. Whether the paint residues concerned are on the surface now or are sandwiched between layers of mineral accretion is in principle irrelevant. Even the presence of a sequence of dates is no guarantee, because if there were conti-nuous contamination of a cumulative stratigraphic sequence the dates might well be in sequence, but they would still be invalid.

With carbon dates from paint residues dominated by charcoal pigment, these considerations are perhaps less paramount, though still applicable. Since such samples consist primarily of charcoal, the impact of distorting contaminations present would be somewhat lessened, but other reservations apply to such dates. The carbon isotope result of charcoal never refers to the event of rock painting, it can at best only indicate the time when the tree from whose wood the charcoal derives assimilated carbon dioxide from the atmosphere (Bednarik 1994c, 1996, 2000a). This is followed by the death of the tree or branch, its oxidation, and finally the selection of the resulting charcoal as pigment. This chain of events may occupy only decades or centuries, but it may equally well take tens of millennia. A documented example of this is provided by David et al. (1999): a dated charcoal inscription at a site in northern Queensland, Australia, apparently of 1894, yielded an age of over 1300 radiocarbon years. There is also the possibility, however remote, that fossil wood had been involved (Schiffer 1986; Fetterman 1996). Much more reliable is the method of analysing soot deposits on a cave ceiling, but this has so far been successful only at one site (Clottes et al. 1995).

Two specialised methods have been developed to process samples from pictograms for carbon nuclide analysis, and to guard against some possible contaminants. One uses a low-temperature, low-pressure oxygen plasma to oxidise the organic matter (Armitage et al. 2000; Chaffee et al. 1993; Ilger et al. 1996; Russ et al. 1990). The second is ‘focused laser extraction of carbon-bearing substances’ (FLECS), followed by accelerator mass spectrometry (AMS) radiocarbon dating (Watchman 1993; Watchman and Lessard 1993). Here, a small sample of a rock art-related substance is combusted with the help of a laser, and the resulting carbon dioxide is then reduced to a graphite target for AMS dating. However, both methods cannot discriminate between different types of organic matter. They are certainly ‘direct dating’ methods in the sense of the term’s definition, but their results are not readily relatable to the actual age of the art concerned. This is because they cannot be targeted on a particular substance, but may in fact provide composite results to which any organic matter present in the sample has contributed. This can include organic pigment or binder, microscopic biota, lipids, proteins, carbohydrates, vegetable remains such as brush fibres, airborne debris and so forth. It can also include natural graphite, whose presence at a rock art site Watchman (1995) has already demonstrated, and which has a significant ageing effect on samples. It is therefore likely that two dating results from different locations on the same motif, using precisely the same technique, will provide different results, depending on the proportions of the contaminating components (see McDonald et al. 1990 for greatly incompatible results from charcoal of a single motif). This renders it even impossible to make precise allowance for the contaminants by determining them and their magnitudes chemically or microscopically from a control sample: their relative proportions may differ locally. Other concerns expressed about the uncritical acceptance by archaeologists of experimental data of this kind are that the petrology and morphology of the rock substrate may be inadequately accounted for; the presence of unknown contaminants; the neglect of 13C measurements; effects of variations in techniques; and that indiscriminate sampling in the course of applications of this immature methodology may not be justified.


Rock art dating – back to the main page

References list for rock art dating