Stratigraphy, Harris matrices & relative dating of Australian rock-art

Rock-art, despite much ingenious effort (e.g., among many, Watchman et al. 1997), remains difficult to date by absolute methods, so relative dating has a central importance much as applied to dirt archaeology in the era before routine radiometric dating. It is sound relative dating which will show just what the entities are to which absolute dates may be connected. The first basis for relative dating is the determination of sequence: what motifs done by which techniques in which materials precede and follow each other; and the first basis for sequence is physical superposition, in which one figure plainly overlies another or in the case of rock-engravings one figure clearly cuts through another. But often figures do not cut or superpose each other so no relation of sequence exists: and sometimes figures are cut through each other without sequence being clear, or are so much overpainted that the older figures are impossible to discern.

Figure 1 shows part of a surface at Garnawala 2, on Innesvale cattle station, in the Wardaman people's country of the northwestern Northern Territory, tropical Australia. Its surface, about 2·7 m high by 1·9 m wide, is painted with a mass of overlapping and superimposed paintings. On it we identify some 32 individual figures, amongst which are 58 observable superpositions involving 31 of the figures. It is a rare instance of a surface with the ideal combination: the figures sufficiently overlap to give sequence, but are sufficiently distinct and exposed to be clearly recognizable.

The Harris matrix method (Harris 1989) was invented as a systematic means to express and to extract all information as to sequence provided by observing relations between stratigraphic contexts in deposits. It can equally be applied to rock-art sequences, as was first done some years ago (Chippindale & Taçon 1993; see Mguni 1997 for a more recent application), and as is being done with this panel. The central white emu on this surface, stratigraphically the most recent figure, is painted in a striking and distinctive manner; Wardaman people remember the man who used to paint large emus so. This and other sufficiently characteristic subjects and manners of depiction, when they recur at sites, enable the Harris matrices from separate sites or panels to be tied together. There is a closely comparable large white emu on a panel at near-by Yingalarri, where again the painting sequence lends itself to a Harris matrix analysis of sequence, and the two large white emus can be equated as likely to be contemporary. Linked together, these few key surfaces can define the Wardaman sequence using a kind of qualitative method. The Garnawala panel has no engravings; at sites where there are both paintings and engravings the chronology of the two techniques can be related.

Alongside these few Wardaman panels showing an intricate sequence are many at each of which there are only a few superpositions. Taken together these provide a mass of observations, so one can begin to work in a quantitative way, leading to a kind of quantitative stratigraphy. Alongside the observation that the big white emu at Garnawala 2 is over polychrome figures of Ancestral Beings, one could say in what proportion of all relations of sequence between white and polychrome painted figures, it is the white which is on top. Amongst older paintings, it can be hard to identify which figure lies on top, as the pigments have 'melded together'; a quantitative approach to large body of observations, finding in it a repeated pattern, is not upset by a few mis-observations.

Wardaman country (Flood & David 1994) lies in a strategic place in Australian rock-art (Flood 1997), central within its great northern province of rock-art with its two sub-provinces of Arnhem Land (northeast of Wardaman country) and the Kimberley (west of Wardaman country). Its chronology is key to that regional pattern, and therefore central to the emerging and extraordinary Australian 'prehistory of the Dreaming' (Mulvaney & Kamminga 1999). Unusually for rock-art we already seem to have reasonably reliable chronologies in northern Australia, at least for Arnhem Land (Chippindale 1998), chronologies which document remarkable change and remarkable continuity. These qualitative and quantitative approaches to sequence are good aids to building those chronologies.

Acknowledgements
We thank the Wardaman Aboriginal Corporation and Wardaman colleagues for their friendship, enthusiasm and access to sites on their land. We thank the University of Cambridge, its McDonald Institute of Archaeology and the Australian Institute of Aboriginal and Torres Strait Islander Studies for funding fieldwork and analysis in 1999­2000.

References

• CHIPPINDALE, C. & P.S.C. Taçon. 1993. Two old painted panels from Kakadu: variation and sequence in Arnhem Land rock art, in J. Steinbring et al. (ed.), Time and space: dating and spatial considerations in rock art research (papers of Symposia F and E, AURA Congress Cairns 1992): 32­56. Melbourne: Australian Rock Art Research Association. Occasional AURA Publication 8.
• FLOOD, J. 1997. Rock art of the Dreamtime: images of ancient Australia. Sydney: Angus & Robertson.
• FLOOD, J. & B. David. 1994. Traditional systems of encoding meaning in Wardaman rock art, Northern Territory, Australia, Artefact 17: 6­22.
• HARRIS, E.C. 1989. Principles of archaeological stratigraphy. 2nd edition. London: Academic Press.
• MGUNI, S. 1997. The evaluation of the superpositioning sequence of painted images to infer relative chronology: Diepkloof Kraal shelter as a case study. Unpublished BA (Hons.) thesis, Department of Archaeology, University of Cape Town.
• MULVANEY, D.J. & J. Kamminga. 1999. Prehistory of Australia. New edition. Melbourne: Melbourne University Press.
• WATCHMAN, Alan L., G.L. Walsh, M.J. Morwood & C. Tuniz. 1997. AMS radiocarbon dating age estimates for early rock paintings in the Kimberley, NW Australia: preliminary results, Rock Art Research 14: 18­26.