Quadruped pictograms in sandstone shelter, central India
Where rock art has become covered by sediment, concealing strata may be considered as postdating the art. They may contain evidence that permits the estimation of the time of deposition, including artefacts presumed datable by their type, organic remains that contain residual radiocarbon, or mineral grains datable by such methods as TL or OSL analysis. While it is not acceptable to define the sediments themselves as being datable by such methods, under favourable conditions such dating information is likely to provide a fair idea of the time of their last deposition. Sediments used in this way must not have been disturbed subsequently, or be the result of deflation or colluvial processes. A recent example of a spectacularly false dating of excavated petroglyphs is provided by a vertical art panel at Fariseu in the Coa valley of Portugal, where it was claimed the petroglyphs must be in excess of 25 000 years old because the deposit contained purported Gravettian artefacts. The colluvium in question was in fact under 17 years old (Abreu and Bednarik 2000) and the assumed age of the perceived artefacts was entirely irrelevant to the age of the petroglyphs.
Worldwide, the minimum dating of rock art by excavation has not been possible in more than a very few instances. Cases of pictograms that have been discovered under archaeological deposits are particularly rare, while the prospects are considerably better for petroglyphs. There have been attempts on most continents to relate pictograms to pigment traces found within a nearby sediment (Macintosh 1965; Wakankar 1983; Combier 1984; Linares 1988; Chippindale and Tacon 1998), but such a link has not been satisfactorily demonstrated in any one instance. A comment sometimes seen is that the ochreous pigment in the ground was found to be of the same hue as the paint on the wall. In reality, iron oxides and hydroxides are highly susceptible to chemical alterations that inevitably involve changes in reflec-tive properties (colour), and it is naive even to assume that either the pigment on the wall or the pigment in the ground would retain the original colour. Both are likely to be modified differently, and will probably not match, bearing in mind the metastable character of all iron salts (which form the majority of pigments used in rock art).
Similarly, there are many claims in the archaeological literature that rock art dating has been achieved because the style or purported content of excavated and roughly dated portable art is identical or similar to that of otherwise undated rock art. None of these claims are listed or discussed here, because they have no scientific or falsifiable basis; they are anecdotal and refer to subjective perception. Besides, even if the presumed similarity were refutable, it would still be very possible that a rock art motif could be copied many millennia after its creation. Hence this kind of 'dating' is not evidence.
It must also be remembered that 'dating' via excavation is an indirect form of dating rock art, i.e. we have to accept inductive pronouncements on trust, such as the validity of the chronological association of charcoal and sediment, the claim that there was no recent contamination in the charcoal sample, and we must accept the archaeologist's pronouncements concerning sedimentological issues. Furthermore, dating of rock art by excavation provides never more than minimum ages, and it offers no evidence of how old the rock art was at the time it became concealed by the sediment. This unknown time span may be as long as a hundred millennia, or as short as a few hours.
Logically there are two different types of processes that may cover rock art by sediment strata. Rock art on either vertical or horizontal panels may be covered in situ, and such an occurrence was first roughly dated by Daleau (1896), at Pair-non-Pair, France, after fourteen years of investigation. Several subsequent minimum datings of petroglyphs by excavation have been reported (Lalanne and Breuil 1911; Lemozi 1920; David 1934; Passemard 1944; Ampoulange and Pintaud 1955; de Saint Mathurin 1975; Anati 1976: 34, 41; Rosenfeld et al. 1981; Cannon and Ricks 1986; Steinbring et al. 1987; Crivelli et al. 1996). Pictograms are very rarely preserved below ground, but they were convincingly minimum-dated through a concealing sediment at Perna 1, Brazil (Bednarik 1989b; Pessis 1999).
Alternatively, a fragment of decorated rock may have exfoliated and fallen to the ground, where it eventually became covered by sediments. Several exfoliated petroglyph fragments have been minimum-dated in this fashion (Capitan et al. 1912; Hale and Tindale 1930; Mulvaney 1969: 176; Thackeray et al. 1981; Fullagar et al. 1996; Roberts et al. 1998), whereas similar claims for exfoliated fragments of pictograms have tended to be less persuasive (except perhaps O'Connor 1995). Dating attempts of excavated petroglyph fragments may well provide a fairly precise timing for the event of exfoliation, which probably occurred very much later than the making of production of the rock art. Such information concerning the minimum age of the rock art offers the potential of using it in calibrating another time-dependent process that relates to the art. For instance, relative weathering or microerosion data from a panel may potentially provide relative age estimates of original petroglyph and exfoliation scar. If the time of exfoliation can be estimated, this may under favourable conditions lead to a fairly reliable age estimation for the rock art. For this or similar approaches to be successful, excavation needs to be combined with a direct dating method, and thus becomes supporting evidence. It is in this kind of framework that archae-ological excavation can assist in securing convincing age estimates of rock art.
It may seem self-evident that mere proximity of rock art to some datable evidence of occupation cannot be adequate proof of the rock art's age by any measure, and yet there are numerous instances on record, from many regions of the world, where archaeologists have deduced the age of rock art from evidence found merely nearby. Of course there may be cases where such precipitate claims were in fact correct, but this is hardly a justification for desperate proposals in the first place, and for presenting such mere guesses as admissible evidence without any corroborating support.
To appreciate the full shallowness of this strategy of dating rock art it is necessary to examine some points of logic. It is based loosely on the intuitive perception that different activity traces of the same period are more likely to occur together than at diverse locations. The concept of activity foci in an 'archaeological space' is thus ignored: the probabi-lity that two types of occupation traces found at one site are contemporaneous is millions of times greater at some randomly selected site in a featureless plain or desert, than it is at a site that was an occupation focus, such as a rockshelter, a cave, a spring or other favoured locality. In other words, if one discovered some petroglyphs together with stone artefacts at some seemingly 'random' locality on the Nullarbor Plain, the world's largest limestone karst, the chances that these two features dated from the same time would be very high indeed. If one discovered these petroglyphs and lithics in Koonalda Cave, a Nullarbor site of highly focused human activity over a long time period, the probability that they represented identical ages would be almost nil. In short, logic tells us that the lumping together of rock art and other evidence of human presence in places that were likely to be much frequented is a practice based on false premises.
A great deal of rock art in the world, possibly most rock art, occurs at localities that can be assumed to have been activity foci for as long as the region has seen human occupation. Rock art usually marks special places in the landscape, it is not distributed randomly in it, and ethnography has often shown why that is so. Similarly, occupation sites are not randomly distributed, their locations, too, are highly predictable. They often coincide with the types of localities favoured in rock art production. Therefore any attempt to extrapolate the dating of any such occupation evidence to the same site's rock art is unacceptable if it is not corroborated by alternative and credible methods.
The issue is best illustrated with an example. An extensive study of the rock art sites of the Koolburra Plateau in north Queensland (Flood 1987) has shown that 48% of these sites have no sediment deposits or traces of occupation, while another 16% have sediment cover over less than a quarter of their floor area. Stone tools were observed on the floor of only 4% of the sites. It follows that the majority of these sites offer no evidence of human presence other than the rock art. If we insisted that rock art must be accompanied by occupation evidence, it would logically follow that most Koolburra rock art should not even exist.
In many cultural contexts rock art production is segregated from domestic activities, and there are even societies who practise exclusion of some members from the cultural practices that result in some forms of rock art. Clearly in such cases it would need to be assumed that utilitarian occupation evidence found at a rock art site was likely to date from a time other than the time the art was made, because there would have been no domestic activity at the site when it served as a (perhaps sacred) ritual site.
A classical example of a claim to have dated rock art by proximity relates to the petroglyphs at Gum Tree Valley near Dampier, Australia. Lorblanchet (1992) secured a single radiocarbon 'date' of about 18 500 years from a large trumpet shell, a surface find from near the petroglyphs, which is over twice as old as all other dates from the site. On this basis alone he then confidently constructs an entire stylistic and formal chronology of the region's massive rock art assemblage spanning more than 18 000 years (Lorblanchet 1992: Figs 20, 21), based on such feeble evidence as subjectively judged patination differences, spatial distribution and perceived subject matter. Although credible dating evidence is still lacking, the assemblage in question appears to be of the Holocene, and all other of Lorblanchet's dates are under 7000 radiocarbon-years. The archaeological occupation evidence consists of shell middens, which implies that the site became settled only when the present sea level was approached, i.e. during the early Holocene.
The practice of dating rock art through the proximity of supposedly datable occupation evidence may be valid occasionally, but on a logical basis it should be expected to provide false deductions much of the time. Statistically it may be as useful as reading tea leaves or animal entrails, which also results in occasionally valid predictions.