Arnhem Land prehistory in landscape, stone and paint.

Tacon, Paul S.C. ; Brockwell, Sally

Arnhem Land at the end of the Pleistocene Making sense of information about the human past is one of the primary goals of archaeological research. In western Arnhem Land we may obtain information from two very different but complimentary sources, shelter walls/ceilings and shelter deposits. The region is well known for its spectacular galleries of rock-art and for its deep and extensive deposits, both of which reach back in time to considerable antiquity. The problem is that on the walls and ceilings we have lots of information from the past but few secure dates. The opposite is true of the deposits where we have sequences of generally reliable dates but very little information. There are few links between the two, and their inter-relationships are not well known. Any links between them are important because they allow better understanding of cultural changes in Arnhem Land that occurred in response to environmental transformations. Through the rock-art we are able to see how some changes were represented by prehistoric Arnhem Landers themselves. Indeed, the rock-painting record is so detailed that a wide variety of material culture and social changes can be outlined in ways that would not be possible if we were to focus solely on the scant material remains.

The Pleistocene - Holocene transgression profoundly affected ancient Arnhem Landers. During this time sea-levels rose, and lands were flooded, and the land-bridge between northern Australia and southern New Guinea was inundated. There was equally dramatic climate change, from extreme aridity to the monsoonal climate known today. During this process, indigenous peoples living throughout northern Australia had to forge new relationships to land, to people and to other species of plants and animals. Adjustments were required to increased rainfall, warmer local climates, flooding, loss of habitats, population shifts, culture contacts, and new sources of food, water and raw materials. The precise responses by specific groups in local areas may never be known, but more general responses can be modelled from evidence in the ground and painted on their rocky surrounds.

Jointly, these records allow us not only to document changes in material culture resulting from the effects of the Pleistocene - Holocene transition but also concomitant social changes. We argue that the fundamentals of Arnhem Land culture can be traced back to the end of the transition and many have their origins in it. These include a simple tool technology, elaborate kinship systems, clan estates that focus on totemic sites in the landscape, a complex belief system that features the Rainbow Serpent as a powerful metaphor and symbol, regional linguistic differentiation, and a detailed visual communication system that revolves around relationships between peoples, Ancestral Beings, other species, time-periods and landscapes.

The environmental sequence

The environment today Kakadu National Park is located 200 km east of Darwin in the `Top End' of the Northern Territory. It is situated about 12 [degrees] south of the equator in a sub-humid savannah environment. In the north it is bounded by the Van Diemen Gulf while beyond its eastern border is the former Arnhem Land reserve. The region under discussion here straddles both Kakadu and the western section of Arnhem Land (Figure 1).

This region contains a number of landforms, including floodplains, lowland plains, sandstone escarpment and the dissected Arnhem Land plateau, and five major rivers, the East Alligator, South Alligator, West Alligator, Wildman and upper Katherine Rivers (Figure 1). The Alligator Rivers floodplains, one of the major freshwater wetlands systems in Australia, include the wetlands associated with the South Alligator River, near-by East Alligator River and one of its tributaries, the Magela Creek. Altogether these wetlands cover an area of about 600 sq. km (Hope et al. 1985: 237).

The climate is markedly seasonal. The dry season from May to December sees virtually no rainfall, with a shorter wet season from December to April.

The environment 15,000 to 7000 b.p. At the beginning of this period sea-level was c. 120-140 below present levels (Chappell & Grindrod 1983b: 67), Australia and New Guinea formed one continent connected by a vast plain extending over the Arafura shelf, from northern Australia to the mountains of New Guinea (Jones & Bowler 1980: 14). The western Arnhem Land sites dated to this period, which are now located on a coastal plain only 50 km from the sea, were then situated about 350 km inland, overlooking deeply incised river valleys (Jones a Bowler 1980. (Figure 2a).

Allen & Barton (1989; 5-12). have reviewed the climatic and geomorphological evidence for environmental change for this period in northern Australia. The climate was drier and cooler than at present, although the seasonal wet/dry regime was already in place. A large brackish lake in what is now the Gulf of Carpentaria had probably contracted at this time, causing a decrease in cyclonic activity and atmospheric disturbances leading to reduced rainfall (Jones & Torgersen 1988; Torgersen et al. 1988. This effect was intensified by rising temperatures which increased evaporation. It is estimated that rainfall in the Alligator Rivers region was reduced by one-half to one-third at this time (Allen & Barton 1989; 5-7), and that the climate and low, open woodland vegetation was similar to the semi-arid zone surrounding present-day Tennant Creek c. 800 km to the south (Jones & Bowler 1980. 14; Nix & Kalma 1972; Allen & Barton 1989:7).

There is evidence of significant temperature rises between i 15,000 and 14,000 years b.p. (Pittock & Salinger 1983: 124). As a result, continental ice sheets started melting and the sea-level began to rise. In northern Australia the effects of this rise were particularly dramatic as the Sahul shelf and the Arafura plain were rapidly flooded between 13,000 and 10,000 b.p. by which time the sea had reached a level 30 - 40 m below present (Chappell & Grindrod 1983b: 67) (Figure 2b). By this time the Kakadu region was possibly only 100 km from the coast.

Subsequent sea-level rise was slower. By 9000 b.p. the sea was c. minus 20 metres, and Torres Strait was severed between c. 8500 and 6000 b.p. By 7000 b.p. the sea was only 4-6, metres below present levels and stabilized c. 6000 b.p. [Chappell & Grindrod 1983c: 87) (Figure 2c). By 7000 b.p. the Arnhem Land plateau was almost as close to the coast as it is today, perhaps only 10 km further south. If 25 to 45 metres of land was being lost each year between 17,000 and 8000 b.p., as suggested by Allen and Barton (1989:7), significant changes for people living close to the coast were occurring in single life-times, although rises may have been intermittent rather than continuous (Chappell 1983: 122: Chappell & Grindrod 1983b. 67).

With the warming of ocean temperatures, precipitation increased, a wetter and perhaps warmer period than at present has been proposed for northern Australia in the period between 10,000 and 7000 b.p. (Ash 1983: 90; Chappell & Grindrod 1983c: Hope 1983: 97 - 8; Jennings 1975: 251-2; Jones & Torgersen 1988; Nix & Kalma 1972; 82-6; Stocker 1971; Torgersen et al. 1988). Other researchers have argued that this period was cooler (Aharon 1983:89; Kershaw 1983: 100-101, Kershaw A Nix 1988. 600; Kershaw, this volume). The amelioration of the climate in terms of precipitation meant that the environment became less harsh. The vegetation was dominated by woodland and open forest, and the wetter conditions encouraged the spread of rainforest (Kershaw 1983; 1985; Stocker 1971).

The various river systems of northern Australia responded differently as sea-levels rose and down-cut river valleys were drowned. Some, such as Darwin Harbour, became deepwater embayments. Others, through processes of sedimentation, formed mangrove swamps and, in the late Holocene, freshwater swamps (Chappell 1988). M@. The rivers of the Kakadu region are examples of the latter type. From 80000 to c. 7000 b.p. the South Alligator River and the Magela Greek went through a `transgressive phase' during which mangroves rapidly invaded newly formed intertidal zones (Chappell 1988:41; Clark et al. 1992. 90; Hope et al. 1985; Woodroffe 1988:3).

The deposit sequence

Two problems when considering the archaeological deposit sequences of Kakadu at the Pleistocen - Holocene transition are. * site settlement patterns for this period are incomplete because sites located along the river margins or on the Arafura shelf are submerged., * deposits that have survived from this period contain only stone artefacts and ochre pieces before 7000 b.p.

Allen & Barton (1989: 103) observed that these gaps in the data might always limit our understanding of Aboriginal life in this period. However we suggest that linking the archaeological deposits and the rock-art of the same period may overcome such limitations.

That there are only seven excavated sites in the Kakadu region whose sequences cover the period of 15,000 to 7000 b.p. also limits our understanding. Also, since deposition rates in some of them are as little as one centimetre per millennium and some have only a few radiocarbon determinations, it is difficult to isolate this period within the sequences and to assess whether occupation is continuous through this period. The sites are Malakunanja II and Nauwalabila I, both older than 18,000 b.p. (Jones & Johnson 1985a; Kamminga & Allen 1973; Roberts et al. 1990c; 1993); Ngarradj Warde Djobkeng, dated to c.8500 b.p. (Allen & Barton 1989; Kamminga & Allen 1973); and Malangangerr and Nawamoyn, dated from the middle of the lowest deposit to between 18,000 and 24,000 b.p. (Schrire 1982). Jimeri Il (Schrire 1982) is dated to c. 7000) b.p. and Nawulandja (Kamminga & Allen 1973) has a basal age of c. 40000 b.p. Another site, Jimeri I, has been dated to c. 4000 b.p. at the base, although another determination of c. 10,000 b.p. from the basal rubble hints at an earlier occupation (Schrire 1982: 152). As the material from this horizon has not been separated from that of the upper levels, this site has not been included in this discussion.

These sites are all located within rock-shelters in outliers of the Arnhem Land escarpment or along its edge (Figure 1). Malakunanja II, Malangangerr, Nawamoyn and Ngarradj Warde Djobkeng are located in outliers which today are close to the flood-plains of Magela Creek and the East Alligator River, at the base of the Arnhem Land escarpment and c. 50 km inland from the present-day coast. Jimeri I and Jimeri II are located in a plateau valley surrounded by escarpments, 30 km east of these sites. Nauwalabila I is located in a plateau valley c. 70 km south of the other sites.

There are various chronological difficulties with some sites. The available radiocarbon determinations do not always match neatly with each other and there is debate over what are the 'real' dates for some sites. For instance at Nauwalabila I, Jones A Johnson (1985a: 181 - 2) rejected a date of c. 12, 000 b.p., accepting instead a date of c. 19,000 b.p. obtained by Kamminga & Allen (1973: 95) for the same approximate level. There has also been some debate over the stratigraphic integrity of Malakunanja II, as to whether artefacts located in deep sandy deposits have been displaced post-depositionally (Allen 1994; Bowdler 1990; Hiscock 1990, Roberts et al. 1990a, 1990b, 1990c., 1994. Conjoined sequences from a site in central Queensland which contains similar deposits demonstrated that some artefacts had moved vertically a maximum of 30 cm, covering a period of 2500 years, despite the deposits having the appearance of undisturbed sediments (Richardson 1992: 417). As several of the sites under discussion here contain deep sandy deposits, this could pose a problem for assigning precise dates to assemblages. Roberts et al. (1990b: 125-6; 1994: 614-15) have dismissed the likelihood of any major downward movement of artefacts at Malakunanja II. the sequences of change in assemblages matched those of other sites in the region; there was no decline in artefacts concentration nor sorting of the artefacts by size or density, there was a pit containing artefacts which was neither intrusive nor the result of post-depositional processes, and the nature of the sediment, although sandy, tends to pack down tightly. Consequently, in the view of these authors, conditions do not favour vertical movements of more than a few centimetres.

With respect to continuity, only 14 cm of deposit accumulated between c. 18,000 and c. 7000 b.p. in Malangangerr; at Nawamoyn 32 cm of deposit was laid down between c. 21,000 and c. 7000 b.p. (Allen a Barton 1989: 85; Schrire 1982). It has been argued that these slow deposition rates may represent a hiatus in occupation (Allen & Barton 1989: 86 Jones & Johnson 1985a: 215; Kamminga & Allen 1973: 25). Schrire (1982:230), on the other hand, argues that occupation at both sites was continuous. At Nauwalabila I, about 100 cm of deposit accumulated between c. 19,000 and c. 6000 b.p. based on the radiocarbon chronology (Jones a johnson 1985a; 181). The rate of accumulation of sand in the site was roughly constant through the late Pleistocene and the early Holocene (Hope et al. 1985: 230-31), and Jones & Johnson (1985: 230-31) argue that Nauwalabila I was continuously occupied through this period.

The sites and their sequences Schrire (1982), from her 1960s investigations of five sites, three on the plain (Nawamoyn, Malangangerr and Paribari) and two in the nearby plateau valleys (Jimeri I and Jimeri II), defined two successive industrial traditions in western Arnhem Land. The earlier one, which she argued lasted from 24,000-20,000 b.p. until 6000-5000 b.p., is characterized by flaked tools, ground axes, hammers, grind-stones and ochre (Schrire 1982: 239); the later one, from c. 5000 b.p., is characterized by unifacial and bifacial points, small rectangular scraper adzes, utilized flakes and edge-ground axes (Schrire 1982: 239, 245, 247-8). Because of the presence of forms like horsehoof cores and steep-edge scrapers Schrire (1982: 254) related the early industry to an Australia-wide Core Tool and Scraper Industry. Most of the tools were manufactured from quartzite with quartz being the other common raw material (Schrire 1982:240), while the edge-ground axes were manufactured from volcanic materials. Several are `waisted' or grooved and some are `stemmed', possibly for hafting. Axes are present through the deposits of all three plains sites, from the Pleistocene to the Holocene periods (Schrire 1982: 241; see also Morwood & Hobbs, this volume). Grind-stones associated with hammers or pounders also occur at all. three sites. Ochre in a wide variety of colours and shapes occurs throughout the sites. A fragment of ochre-stained schist from Jimeri II suggests that ochre was crushed at that site (Schrire 1982:241). Schrire (1982:242) concluded that little could be said about the identity or behaviour of the people, who used the shelters briefly and discarded little. While the axes suggested cultural continuity, there were too few flaked implements to show specific changes through time.

Subsequent work by Kamminga & Allen (1973) generally confirmed Schrire's two-phase sequence for western Arnhem Land, but revealed some differences in the nature of the early industry at Ngarradj Warde Djobkeng, Malakunanja Il and Nauwalabila I on the one hand, and Nawamoyn and Malangangerr on the other (Allen & Barton 1987: 21). At Nauwalabila I, Kamminga & Allen demonstrated that the lower industry is characterized by unmodified flakes made predominantly from quartz, except for one period where chert was dominant and generally lacked diagnostic implements. This industry was dated between c. 3000 and c. 19,000 b.p. Further investigations at Nauwalabila by Jones A johnson (1985a) confirmed the earlier results and provided extra details. They concluded that the quartz and chert assemblages between 6000 and c. (19, 000 b.p. consisted mainly of utilized flakes, cores and ad hoc, scraper edges (Allen & Barton 1989: 79; Jones & Johnson 1985a:215). At the base of the deposit the diagnostic tools were described as being `steep-edge scrapers or horsehoof-core tool fragments', estimated then to date from c. 19,000 to perhaps 25,000 b.p., similar in age to those found at Nawamoyn and Malangangerr (Jones & Johnson 1985a: 214-15 Recent optically-stimulated luminescence dating at Nauwalabila I may make these artefacts considerably older (Roberts et al. 1983), a matter for some debate (Allen 1984; Allen & Holdaway 1985; Roberts et al. 1994).

At Malakunanja II quartz flakes dominate the assemblage. Only two definite stone implements, both points, were recovered from this excavation both in the upper layers. There are no discernible changes in raw materials or artefact types in the lower layers, dated to between c. 4000 and c. 18,000 b.p., despite marked stratigraphic changes (Allen & Barton 1989: 80-81) At Nawulandja, quartz fragments occur throughout the deposit and form over 51% of the assemblage in the lower layers dated to c. 8500 b.p. The only diagnostic implements, points, may date to only c. 1500 years b.p. (Allen & Barton 1989: 81-8).

Ngarradj Warde Djobkeng has been dated to c. 8500 b.p. at a level about a third of the way through the deposit. Given the oldest dates for the near-by sites of Matakunanja II, Nawamoyn and Malangangerr, the base of Ngarradj may well be older than 25,000 b.p. (Allen A Barton 1989: 29-30). There were changes in flake size and dominant raw materials before woo b.p. Allen & Barton (1989: 74) conclude that the evidence may indicate a change in flaking methods or technology through the lowest levels at Ngarradj, but that the data are insufficient to define an assemblage for the basal deposit. Unlike Schrire (1982), who argues that the stone artefacts in the lowest levels at Nawamoyn and Malangangerr were the waste products of in situ manufacture of stone implements, Allen & Barton (1989: 37) conclude that, given the lack of large implements, the major industrial activity before 8500 b.p. at Ngarradj was the production of untrimmed medium quartzite and quartz flakes, which were probably used as implements. Quartz cores dominated the lower assemblages, which contained three times as many cores as the layers above (Allen & Barton 1989:49).

Jones & Johnson (1985a: 215-16) suggest that the Core Tool and Scraper Tradition identified in the lower levels of sites in Kakadu may be confined to the period before 18,000 b.p.; the sequences at Malangangerr and Nawamoyn may be truncated, with a hiatus of occupation between 18,000 and c. 6000 b.p. when the later industry of small tools appears. This would explain the differences between the older assemblages at these sites and those of Ngarradi, Nauwalabila I and Malakunanja II and the hint of a different industry in the lowest level at Ngarradj. If such is the case, a pattern begins to emerge for the Pleistocene-Holocene transition period of a stone industry comprising amorphous quartz and quartzite flakes, both readily available local raw materials. However, more precise dating of different levels in the sites is currently required to demonstrate this.

In summary, a meagre number of sequences in western Arnhem Land span the Pleistocene-Holocene transition; low numbers of radiometric dates for the relevant parts of these sequences, together with a lack of diagnostic artefacts which might link these sites, makes it uncertain which of these sequences most concern it, and exactly how they inter-relate. All the sites are rock-shelters which only contain stone artefacts, ochre and some charcoal for the period under review so that no subsistence evidence is available. The stone tools are amorphous and made with local raw materials. Given this unprepossessing depositional record, can we rectify it with alternate data sources?

The rock-art sequence

During the past two decades a number of rock-art sequences have been proposed by a wide variety of researchers. These include Brandl (1973), Chaloupka (1977; 1984a; 1993a), Haskovec (1992), Lewis (1988), Tacon (1987; 1989b; 1982), and Tacon & Chippindale (Chippindale & Tacon 1993; Tacon & Chippindale 1994). There has been much debate about the various terms used to describe the rock-art in Arnhem Land shelters (e.g. Chaloupka 1994) but the fundamentals of the manners of depiction, different periods, the sequence of changes, and the nature of particular forms of representation are essentially similar. Here we review the general sequence without reference to small differences or the lively debate that has occurred between many researchers.

The nature of the early tradition

In Arnhem Land the bases of archaeological deposits, containing stone tools and ground pieces of haematite, have been dated to 30,000 b.p. by radiocarbon and over 50,000 b.p. by luminescence techniques. At large sites with considerable depth, such as Malakunanja II and Nauwalabila I, used fragments of ochre have been found at almost every level of deposits (Jones 1985. 297; Smith 1990. While this pigment may have been used for rock-painting, it may also have resulted from pigment preparation for body- or object-painting or other purposes.

Across northern Australia there have been more changes in the forms and styles of rock-painting over time than elsewhere in the world, but rock-art production is unlikely to have occurred at a steady, constant rate over many millennia. Various bursts of art production over time are more probable than a continuous output, and this is supported by an analysis of the frequency of pigment in different levels of deposits at particular sites (see David et al. 1994, and below). Two of the more prolific and significant eras in Arnhem Land are the intervals associated with Chaloupka's (1984b) 'Dynamic Figures, and the more recent `Complete Figure, period (see Tacon 1989b: 141), although the intervening 'Simple Figures/Yam Figures' stage may also have been one of relatively high art production.

Previous researchers have used form, perceived intellectual intent, artefacts illustrated and the environment to divide Arnhem Land rock-art into phases. All approaches have inherent problems. Here we follow Tacon a Chippindale (1994), but many terms coined by Chaloupka (1977; 1984a) are also employed. Particular 'styles', forms, elements and motifs were placed in the appropriate phase after a detailed analysis of superimpositions at two large key sites (Chippindale & Tacon 1993) and at hundreds of smaller sites studied by Chaloupka, Chippindale, Lewis, ourselves and others.

The earliest surviving art forms in Arnhem Land may be engravings of bird, macropod and human tracks, cup-like depressions, grooves and, occasionally, circles. These possible Pleistocene sites are rare but when found in association with rock-paintings they consistently lie underneath them (Sullivan 1988). Other clusters of these engraved motifs are covered with thick silica skins, the most easterly recently described from near the Mann River in central Arnhem Land (Tacon 1993a; 1994b). As yet, there have been no attempts to date engraved motifs but there is circumstantial evidence from elsewhere to suggest they pre-date the transition to the Holocene (see Nobbs & Dorn 1988; 1993a, Rosenfeld et al. 1981).

Chaloupka (1977; 1984a; 1993a; 1993b) argues that the earliest paintings consist of hand and object prints, followed by large naturalistic depictions of animals and humans. Macropods are frequent (Figure 3) but snakes, freshwater crocodiles and a few other species are also illustrated. Chaloupka grouped these representations as the Large Naturalistic Style, but the definition is inexact. Life-size naturalistic portrayals from any period could be placed in this grouping. Style elements peculiar to this body of art need better description. Haskovec (1992) and Tacon & Chippindale (1994) argue that depictions of humans are extremely rare in this early painted style and that animals predominate. These images may date to near the beginning of the Pleistocene-Holocene transition (Tacon & Chippindale 1994).

Dynamic Figures and the Pleistocene-Holocene transition

The succeeding period focussed more on human figures than on large naturalistic animals. Because many portrayals suggest movement and action, Chaloupka (1977) labelled them `Dynamic Figures' (Figure 4). We contend that these figures, plus the complex of `Dynamic' animals and associated art forms, are the oldest that can confidently be placed in the Pleistocene-Holocene transition period. An analysis of the extinct fauna, environments, and material culture depicted, and the dating of silica crusts formed over Dynamic Figures, both suggest an age of at least 10,000 b.p. (Watchman 1987; 1990; Tacon & Chippindale 1994). The paintings reflect an arid environment, not prevalent since the marine transgression. Male figures predominate, often depicted with ceremonial head-dresses and ornamentation, and the material culture, characterized by boomerangs and single-pronged spears, is very different from that of the recent era. Macropods are the most frequently represented animal. In contrast, in more recent rock-art complexes, such as that associated with X-ray art, there is a diversified technology that does not include boomerangs; and fish dominate (Tacon 1989a).

Dynamic Figure paintings are homogenous throughout western Arnhem Land in their content, colour and form, with only small differences noted between central and western Arnhem Land (Tacon 1993a; 1994b). Layton (1991: 165) has suggested that rock-art displaying a regional uniformity of themes, but concentrated at certain localities, reflects the emergence of collective hunting strategies at predictable locations. More regionalism in the art that post-dates Dynamic Figures (Lewis 1988; Tacon 1993b; Tacon & Chippindale 1994), has led Layton (1991; 1992; in press) and others to suggest that changes in hunting strategies and relationships between groups came about as a result of the Pleistocene-Holocene transition.

Dynamic Figures are shown in combat as well as in a variety of hunting and domestic `scenes'. Figures are shown flailing boomerangs, dodging spears and chasing each other with weapons raised. Sometimes animal-headed beings are shown chasing them (see Chippindale & Tacon 1993: 52). Chaloupka (1984b) & Lewis (1988) noted Dynamic fighting scenes but did not particularly investigate them. However, these are among the earliest examples of warrior figures anywhere, and hunter-gatherer warrior art is generally rare. Depictions of fighting in more recent styles have also been identified by Tacon & Chippindale (1994), who argue that they reflect changes in society that occurred alongside the extreme environmental changes of the Pleistocene-Holocene transition.

Change is the theme of subsequent western Arnhem Land rock-art. Human figures became less animated and detailed (Chaloupka's 'Post Dynamic Figures') until limbs and bodies were represented by thin lines, and heads/head-dresses were abstracted into circles, concentric circles, elliptical shapes or a variety of head-dress-like forms (Lewis 1988; Tacon 1993a). Chaloupka (177; 1984a; 1993a; 1993b) labelled these `Simple Figures with boomerangs' and Lewis (1988) noted some of their variability across Arnhem Land. Tacon (1993a; 1994b) and Tacon & Chippindale (1994) noted further regional variability in these forms and the consistency of certain attributes, such as concentric circle heads/head-dresses, across the Arnhem Land plateau. Important in this context are the changes in modes of depiction that occurred during the Pleistocene-Holocene transition and the increased variability within and between regions that developed. Other changes include the eventual disappearance of the boomerang and new weapons, such as `hooked sticks', spear-throwers and multipronged spears (Lewis 1988; Tacon & Chippindale 1994). At some sites human figures were arranged in great battle scenes, suggesting the rise of more tightly organized societies and competition over resources (Tacon & Chippindale 1994).

Subsequent changes in the rock-art

Among the styles that followed the production of Dynamic Figures, Chaloupka's (1977; 1984a) 'Yam Figure' style may have had a mythological function. It is characterized by figures with yam-like, bodies and includes both human and animal-like yams' (Figure 5), as well as some more naturalistic humans and animals. Often they are associated with paintings of flying foxes, birds, and animals with unnatural attributes. Composite Rainbow Serpents (Figure 6) appeared for the first time and soon became common (Tacon, Wilson & Chippindale 1995). Lewis (1988) argues they are a significant unifying symbol between diverse groups of people affected by the disruptions associated with the end of the Pleistocene. Yams, water lilies, and a variety of other plant motifs become commonplace in the art, suggesting a shift in interest towards certain plant foods.

Direct dating of rock-art

The later sequence in Arnhem Land art falls outside the scope of this paper. Relevant in it is the recent dating of figures made in beeswax, a rare material for Australian rock-art which is uniquely amenable to direct carbon-dates (Nelson et al. 1995). The carbon dates place a 'simple X-ray' turtle motif at 4000 b.p., an age consistent with the dates proposed here for elements older in the well-established relative chronology of Arnhem Land rock-art.

Haematite and ochre fragments

Ochre is found throughout Arnhem Land deposits but the significance of variation in ochre deposition has not been adequately explored. In the Victoria River region to the west, David et al. (1991; 377; 1994) have shown that ochre deposition at rock-art sites indicates that occupation of the sites and the appearance or intensification of rock-painting activity in them did not coincide, with painting being a late phenomenon.

The best record of ochre deposition in Arnhem Land comes from Nauwalabila I, Deaf Adder Gorge (Figure 7). Although ochre fragments were found throughout most of the sequence, peaks of pigment deposition can be isolated from tables published by Jones & Johnson (1985a: 186-7, table 9: 4; 220-1, table 9: 13). There are small peaks between the present and 2000 b.p., at 2000 b.p., between 3000 b.p and 4000 b.p., and at 6000 b.p. There are three further peaks between 6000 b.p. and 12,000 b.p. and one just before 12,000 b.p. There was only incidental usage of ochre before this, but fragments were found intermittently down to the bottom of the deposit. From this it might be inferred that in this site Dynamic Figure art, associated with a prolific period of production, should be no older than 12,000-13,000 b.p., when the first peak of ochre occurs. However, Dynamic paintings may equally be younger and associated with the peaks between 12,000 and 6000 b.p. Other forms of supporting evidence are thus needed before an inference from the ochre deposition record can be generally accepted.

Stone tools and flakes

The most commonly depicted stone tool in Arnhem Land rock-art is the hafted axe (Figures 8). Often it appears to have ground edges, and it is frequently shown in use. The earliest examples can be found in Dynamic Figure art Chaloupka 1993a: 120; 1984b: 102, 169, 378, 390), but some later figures were also painted holding them (Figure 9). Lewis (1988: 273) notes a stencil of a hafted stone axe associated with two Yam Rainbow Serpents. Further instances are found in subsequent manners of depiction (e.g. Brandl 1973: 62, 63; Chaloupka 1993a: 57; Tacon 1992: 16). Stone axe-heads, blanks and flakes of volcanic rock, including many with ground edges, are found throughout deposit sequences in Arnhem Land dated to between the late Pleistocene and the present (Jones & Johnson 1985a: 216-18, Schrire 1982). As a consequence, they are not diagnostic artefacts that allow us to date a specific form of rock-art. However, they do give us an indication of the possible antiquity of the earliest depictions.

Schrire (1982) dated deposits overlying the earliest edge-ground axes she recovered from the Malangangerr and Nawamoyn sites to between 18,000 and 23,000 b.p. Jones & Johnson (1985a: 217-18) presented supportive evidence from Nauwalabila I, where axe fragments were found throughout the sequence, from the present to c. 30,000 b.p. However, most were recovered from depths of up to 1.5 metres and were associated with deposits younger than 15,000-16,000 b.p. (Jones & Johnson 1985a; 216).

On this basis we contend that the early and common portrayals of hafted edge-ground axes in Dynamic Figure art are less than 15,000-16,000 b.p., and that the whole Dynamic Figure complex should not be older than this. This accords well with the first peak in ochre fragments in deposits at just over 12,000 b.p., and the two lines of evidence taken together strongly suggest that Dynamic Figure rock-art in all probability is not much older than 12,000 years of age and may be more recent. Significantly, the imagery shows human figures in an environment different from that of the Holocene, one presumed to be much more arid. The gradual disappearance in the art of depictions of boomerangs, previously discussed, reflects change from an area with open grassland to a forested region where these weapons were less effective.

At the other end of the spectrum, long quartzite 'lauwk' blades also appear in some rock-paintings (see Tacon 1991; 203; figure 15); these are only found in the uppermost parts of deposits and are thought not to have been made for more than 1000 years (Allen 1989). Importantly, in the rock-art they are only associated with recent styles such as X-ray paintings.

In between, there is only one important diagnostic stone tool occasionally depicted in Arnhem land rock-art. This is a hafted stone chisel, an artefact Jones & Johnson (1985a: 217) argue was introduced about 4000 b.p. The clearest example of this artefact in the rock-art comes from a site on the plateau, above Twin Falls. It consists of a Simple Figure style human man holding a hafted chisel and a boomerang in its left hand and a barbed spear extending down from what appears to be a spear-thrower in its right hand (Figures 10 & 11). The figure has a large head-dress and is associated with a second figure that carries a boomerang, hooked stick/spear-thrower and two barbed spears. Both figures appear to have grass skirts tied at the waist, and their general form is similar to dozens of comparable figures spread across the Arnhem Land plateau.

The presence of the hafted chisel tells us some Simple Figure style paintings were made more recently than c. 4000 b.p. By association, we can also infer that boomerangs (Figure 10) were still in use after c. 4000 b.p. This provides an important bridge between the ground deposits and those of shelter walls where the environment has precluded the survival of organic object in deposits.

A multi-disciplinary analysis

While the archaeology of terminal Pleistocene coast-lines is now under water, excellent geomorphological data and well-preserved archaeological deposits in western Arnhem Land have allowed reconstructions of human interactions with this changing set of microenvironments at its end-point, after 7000 b.p. The formation of the estuarine and then freshwater wetlands is reflected in the archaeology of the mid- to late-holocene period in terms of open and shelter site distributions and levels of activity in the rock-shelter sites. A very brief summary follows, derived from the principal sources (Allen 1987; 1989; Allen & Barton 1989; Brockwell 1989; Jones & Johnson 1985a; 1985b, Meehan et al. 1985; Schrire 1982; Woodroffe et al. 1986, 1988).

From 7000 b.p., sites in the north clustered along the edges of the Magela Creek and East Alligator River floodplains (Figure 1). Shell middens in the plains rock-shelters of Malangangerr, Malakunanja II and Nawamoyn indicate that, from 7000 to 3000 b.p., one focus of subsistence strategies was estuarine shellfish. The same was the case for Ngarradj Warde Djobkeng from 3500 b.p. and the Paribari rock-shelter, adjacent to the Magela Creek flood-plains, which was occupied for the first time c. 3000 b.p. This evidence correlates with the geomorphological data for widespread mangrove swamp conditions on the floodplains at that time. From 6000 years b.p. there was increased activity at the outlier site of Anbangbang I, which lies on the plains about midway between the wetlands of the South Alligator River and the Arnhem Land escarpment (Figure 1). This may reflect increased occupation of the plains. At the southern escarpment site of Nauwalabila I, activity declined after 6000 b.p., perhaps reflecting a population drift towards the plains with the new estuarine resources.

Following the retreat of the mangroves towards the river margins and the emergence of saline mudflats on the floodplains after 3000 b.p., some sites were abandoned and open sites closer to mangrove zones were occupied on the floodplains themselves. About 3000 b.p. Nauwalabila I experienced increased activity which may reflect a retreat of population to the plateau valleys following the decline of estuarine resources on the floodplains. A similar argument might be made for Jimeri I and Jimeri II.

After 1500 b.p. freshwater wetlands formed in the region and open sites along the floodplain margins of the South Alligator River were occupied. In the south, the plains outlier site of Yiboiog (Figure 1) was occupied for the first time c. 1000 b.p. and there was increased activity at the near-by site of Anbangbang I. In the north, the upper levels of Paribari were dominated by freshwater shellfish. Previously abandoned sites on the northern floodplains may have been re-occupied at this time. At the escarpment valley sites, both north and south, there was a decline in activity which may again reflect the occupants being attracted to the resource-rich freshwater swamps of the floodplains.

We view this as a patterned and, after 8000 years of practice, a well-honed set of human responses to changing environmental circumstances, chief among them being coastal inundation at an average of 40 metres per year, even if sea-rise was not regular.

From this perspective, we can look backwards; with it, a model of coast-hinterland-inland interactions can be constructed for the Pleistocene-Holocene transition, based on site location, occupancy/absence and differing intensities of site use. It is particularly useful to look at the evidence from Nauwalabila I in this context. Given that this site was regularly occupied throughout this period and is located too far inland to be affected by base-level changes induced by sea-level rise (Hose et al. 1985: 231) it makes a barometer for events which occurred elsewhere in the region.

15,000 to 13,000 b.p. At c. 15,000 b.p., population was probably thinly spread in the Alligator Rivers region, with the majority living closer to the coast on the savannah edges of the Arafura plain (Jones & Bowler 1980:14). It is likely that what sparse settlement there was in the then arid inland region of Kakadu was focussed to some extent on the rivers, and we could expect that open sites occurred in the down-cut river valleys.

At this time the outlier rock-shelter sites of Ngarradj Warde Djobkeng and Malakunanja II may have been occupied (Allen & Barton 1989: 102), while the near-by plains rock-shelters of Malangangerr and Nawamoyn had probably been abandoned at the height of the Last Glacial Maximum (Allen & Barton 1989: 84-5; Jones 1985: 295-6; Kamminga & Allen 1973: 25: Schrire 1982: 84-5). Jones (1985: 295) argues that the southern plateau valley site of Nauwalabila I was continuously occupied through this period,

Thus the extremely arid conditions of the time probably led to the abandonment of sites and concentrated populations along the river valleys and in a few outlier sites. Subsistence strategies were probably based on the rivers and creeks and adjacent low open woodland. Importantly, the distribution of the oldest large naturalistic animal art sites closely follows this distribution, for the most part the depictions are of macropods and large pythons - significant food animals associated with an arid climate. People were painting near where they were camping and hunting, rather than in more isolated areas. The body of art is homogenous in subject-matter, form and manner of depiction. This suggests a very open, public art and symbolic system.

There is no suggestion that the encroaching coast, still several hundred kilometres north, had had any impact in terms of population pressure at this time.

13,000 to 10,000 b.p. During this period sea-level rise was rapid, by 10,000 b.p. the sea was only 35-40 m lower than today, with the coast located only c. 100 km to the north of Kakadu. Any populations located closer to the coast of the Sahul shelf would have been in retreat. By this time the climate was also milder, providing a more productive and comfortable environment. In the Kakadu region itself the sea had not yet encroached upon the river valleys. It is likely that the rock-shelter sites of Ngarradj Warde Djobkeng, Malakunanja II and Nauwalabila I were still occupied throughout this period.

At Nauwalabila I the density of site usage has been estimated by calculating the number of stone artefacts per kilogram of deposit. Several peaks occur throughout the sequence (Jones & Johnson 1985a: 183, 186) of which one occurred between 13,000 and 10,000 b.p. To the north, at Ngarradj Warde Djobkeng, there is an increase in the density of stone artefacts before 8500 b.p. in the unit above the basal level (Allen A Barton 1989:27) suggesting an increased use of the site, that might be contemporaneous with that at Nauwalabila I.

The slight evidence available for this period suggests increasing use of sites, which we propose was a result of the marine transgression pushing people from the flooded land into the Kakadu region. The milder climate and more productive environment would have gone some way to ameliorate any pressures caused by putative extra population, but this does not eliminate the possibility of conflict between new groups and the resident population, as suggested by analyses of Dynamic Figure rockart (Chaloupka 1994, Lewis 1997; Tagon & Chippindale 1994). New groups may also have introduced different technologies and cultural practices. As the coast was closer, there may have been increased interaction between coastal and inland populations. Stone resources from the plateau valley quarries and the Arnhem Land escarpment would have become important if stone sources to the north were inundated. Subsistence strategies probably continued to focus on riverine resources and those of the adjacent woodland and open forest. If different technologies were required to exploit the changing environment this is not reflected by the stone assemblages or seen in the rock-art.

10,000 to7000 b.p. Although the sea-level rise had slowed, by 7000 b.p. the sea was 4-6 below its present level, the down-cut river valleys of the region were flooded, and there were large mangrove swamps on the floodplains. Increased rainfall encouraged the spread of rainforest, although open woodland and forest were the dominant vegetation types on the plains.

Two sites, Jimeri I (Schrire 1982: 152) and (Kamminga & Allen 1973: 64-9), may have been occupied for the first time at c. 9000 - 10,000 b.p. Ngarradj, Nauwalabila I and Malakunanja II continued to be occupied (Allen & Barton 1989: Jones & Johnson 1985a; Kamminga & Allen 1973). By c. 6000-7000 b.p. Malangangerr and Nawamoyn were re-occupied (Allen & Barton 1989: 102),, there was increased activity at the southem plateau valley site of Nauwalabila I (Jones & Johnson 1985a: 179, 183, 186); and the northern plateau valley site of Jimeri Il was also occupied (Schrire 1982: 230). Malakunanja II, Malangangerr, and Nawamoyn are all located close to the flood-plains, newly available estuarine resources are reflected in the contents of the shell middens which began to build up from 7000 b.p. Beginning also at this time is a change of emphasis from terrestrial to estuarine in the faunal subject-matter of the rock-art, as well as increasing regional heterogeneity (Lewis 1988; Tagon 1993b; 1994a).

Although rates of change were undoubtedly slow, people made substantial changes to subsistence strategies and probably social organization to meet the demands imposed by a new landscape. Population increase can be assumed as available land was reduced by the flooding of the river valleys and the retreat of the coastline towards the barrier of the Arnhem Land escarpment. The occupation of new sites and increased activity, especially in the more marginal areas of the plateau valleys and the outliers to the south, support this interpretation. Human seasonal movement may also have been accentuated, especially in the north where the floodplains reach almost to the base of the Arnhem Land escarpment, and wet-season flooding would have forced people on to higher ground. There may have been a period of increased conflict based on competition for resources, and marking of territorial boundaries. The earliest battle scenes in the art date to the end of this period (Tagon & Chippindale 1994). At the same time, the emergence of resource-rich estuarine swamps c. 7000 b.p. and of other new resource bases such as rainforests, improved the region,s capability to support an increased population. This situation is reflected by re-occupation and increased activity at sites close to these resources.

Notably, ancestral forms of all the essential elements of the most recent rock-art complex can be found in an earlier body of art thought to date between 6000 b.p. and 4000 b.p. This body of art, which developed into the recent complex prominent for over 3000 years, grew out of environmental, technological and cultural changes that were set in motion by the Pleistocen-Holocene transgression. We see that many key elements of traditional Arnhem Land culture at the time of European contact grow from events which occurred at the end of the Pleistocene after a transition which built on older patterns.

Conclusion

Deposited evidence from 15,000 to c. 7000 b.p. in western Arnhem Land is notable for its lack of diagnostic artefacts. It shows little change in form or raw material throughout, despite some marked stratigraphic changes. Yet the art shows marked changes towards the end of this period. Archaeologically, the intervals bracketing this period are easier to define: a Core Tool and Scraper Tradition is older than c. 18,000 b.p., and a Small Tool Tradition with points and small scraper adzes post-dates 7000 b.p. A hint of more extensive settlement of the region, post-dating 13,000 b.p., accords with a significant increase in art production at about this time. The art of this period supplies some of the missing details regarding the life-style and subsistence patterns of Arnhem Land peoples during the Pleistocen-Holocene transition which the deposited sequences cannot.

The full sequence of art, as well as new subject-matter such as Rainbow Serpents and hooked sticks, needs better documentation in order to provide further insight into ways the end of the transition affected Arnhem Landers culturally. The terminal phase of the transition may have been the most important (Tagon & Chippindale 1994:227):

Indeed, it appears that the rock-art, stone tool technology and social organization that developed during this period laid the foundation for the present: a situation consisting of extreme social and artistic complexity complimented with technological simplicity, a high degree of linguistic diversity and a land tenure system that revolves around clan totemic relationships to particular geological and hydrological centres or focal points within the landscape.

The challenge now is to secure more links between the information on Arnhem Land shelter walls with that recovered from deposits. In this period of environmental transformation, there were many forms of response: in western Arnhem Land symbolic change may have been among the most significant.

Under the impetus of a changing environment the western Arnhem Land response has been to incorporate new ideas, technologies, food-gathering strategies and so forth, so that the culture has become more layered and complex over time. This is particularly reflected in changes to rock-art subject matter. The earliest art, of large naturalistic fauna, focussed on a world of animals. With the rise of Dynamic Figures people began to paint themselves into the pictures, and animals frequently were shown in some sort of human context or relationship. The first mythological, composite beings were portrayed, human-like figures with macropod or flying fox heads. Subsequently, interest shifted to plants, the development of composite Rainbow Serpents that have yam and animal features, yam-like humans and animals, and a range of plant motifs. At the same time, people began to show not only the external features of creatures but also the internal, leading eventually to the life-size polychrome X-ray paintings for which the region is famous. Throughout the Pleistocene-Holocene transition, a human world and then a plant world was added to an original symbol system drawn from the animal world and, at the same time, combinations of elements from different creatures were increasingly made to depict the supernatural.

This continually changing and growing symbol system has operated on a principle of incorporation/inclusion rather than rejection/exclusion for at least 15,000 years. At the same time, the rock-art increasingly became more regionally distinct. Significantly, the process continues in Arnhem Land today as Macassan and European symbols, technologies, ideas, languages, animals and art forms have been and continue to be added to the breadth of indigenous Arnhem Land culture.

References

Aharon, P. 1983. The [O.sup.18] and [C.sup.13] isotope enrichment in the post-Glacial reef complex I from Papua New Guinea and its interpretation, in Chappell & Grindrod (ed.) 1983a: 89. Allen, H. 1987. Holocene mangroves and middens in northern Australia and southeast Asia, Indo-Pacific Prehistory Assoaciation Bulletin 7:1-16. 1989. Late Pleistocene and Holocene settlement patterns and environment, Kakadu, Northern Territory, Australia, Indo-Pacific Prehistory Association Bulletin 9:92-117. Allen, H. a G. Barton. 1989. Ngarradj Warde Djobkeng: White Cockatoo Dreaming and the prehistory of Kakadu. Sydney: Oceania Publications. Oceania Monograph 37. Allen, J. 1994. Radiocarbon determinations, luminescence dating and Australian archaeology, Antiquity 68:339-43. Allen, J. & S. Holdaway. 1995. The contamination of Pleistocene radiocarbon determinations in Australia, Antiquity 69: 101-12 Ash, J. 1993. Rainfall patterns in northeastern Queensland 7[+ or -] ka, in Chappell & Grindrod (ed.) 1983a: 90. Bowdler, S. 1990. 50,000 year old site in Australia - is it really that old?, Australian Archaeology 31:93. Berandl, E. 1973. Australian Abnormal paintings in western and central Arnhem Land.. temporal sequences and elements of style in Cadell River and Deaf Adder Creek art. Canberra: Australian Institute of Aboriginal Studies. Brockwell, C.J. 1989. Archaeological investigations of the Kakadu Wetlands, Northern Australia. Unpublished MA thesis, Australian National University. Chaloupka, G. 1977. Aspects of the chronology and schematization of two prehistoric sites on the Arnhem Land Plateau, in P.J. Ucko (ed.), Form in indigenous art.. schematization in the art of Aboriginal Australia and prehistoric Europe: 243-59. Canberra: Australian Institute of Aboriginal Studies. 1984a. From palaeoart to casual paintings the chronological sequence of Arnhem Land plateau rock art. Darwin: Northern Territory Museum of Arts and Sciences. Monograph Series 1. 1984b. Rock art of the Arnhem Land plateau: paintings of the Dynamic Figure style. Darwin: Northern Territory Museum of Arts and Sciences. 1993a. Journey in time. Sydney: Reed Books. 1993b. You gotta have style, in M. Lorblanchet &, PG. Bahn (ed.), Rock art studies: the post-stylistic era or where do we go from here?: 77-98. Oxford: Oxbow. 1994. Comments on Taron & Chippindale [1994], Cambridge Archaeological Journal 4:240-41. Chappell, J. 1983. Sea-level changes, 0 to 40 ka, in Chappell & Grindrod (ed.), 1983a:121-2. 1988. Geomorphic dynamics and evolution of tidal river and floodplain systems in northern Australia, in Wade-Marshall & Loveday (ed.) 34-57. Chappell J. & A. Grindord (ed.). 1983a. CLIMANZ. Canberra: Department of Biogeography and Geomorphology, RSPacS, Australian National University. 1983b. 15-10ka review, in Chappell a Grindrod (ed.) 1983a: 67-9. 1983c 78[+ or -]a `spike' review, in Chappell & Grindrod (ed.), 1983A:87-8. Chippindale C. g, P. Tacon. 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:32-56. Melbourne: Australian Rock Art Research Association. Clark, R.L. et al. 1992. Late Quaternary evolution of the Magela plain, in R.J. Wasson (ed.), Modern sedimentation and late Quaternary evolution of the Magela Creek plain: 81-157. Canberra: Australian Government Publishing Service. David, B., I. Mcniven A J. Flood. 1991. Archaeological excavations at Yiwarlarlay I: site report, Memoirs of the Queensland Museum 30(3):373-80. David, B. et al. 1994. Of Lightning Brothers and White Cockatoos: dating the antiquity of signifying systems in the Northern Territory, Australia, Antiquity 68:241-51. Haskovec, I. 1992. Mt Gilruth revisited, Archaeology in Oceania 27(2):61-74. Hiscock, P. 1990. How old are the artefacts in Malakunanja II? Archaeologyin Oceania 25:122-4. Hope, G. 1983. Montane Papua New Guinea 9000-5,000 b.p., in Chappell & Grindrod (ed.) 1983a:97-8. Hope, G., P.J. Hughes &, J. Russell-Smith. 1985. Geomorphological fieldwork and the evolution of the landscape of Kakadu National Park, in Jones (ed.) 229-40. Jennings, J.N. 1975. Desert dunes and estuarine fill in the Fitzroy estuary, northwestern Australia, Catena 2:215-62. Jones, M.R. &, T. Torgersen 1988. Late Quaternary evolution of Lake Carpentaria on the Australian-New Guinea continental shelf, Australian Journal of Earth Science 35:313-24. Jones, R. (ed.). 1985 Archaeological research in Kakadu National Park. Canberra: Australian National Parks and Wildlife Service. Special Publication 13. Jones, R. 1985. Archaeological conclusions, in Jones (ed.): 291-8. Jones, R. & J. Bowler. 1980. Struggle for the savanna: northern Australia in ecological and prehistoric perspective, in R. Jones (ed.), Northern Australia: options and implications: 3-31. Canberra: RSPacS, Australian National University. Jones, R. A I. Johnson. 1985a. Deaf Adder Gorge: Lindner Site, Nauwalabila I, in Jones (ed.):165-227. 1985b. Rockshelter excavations: Nourlangie and Mt Brockman massifs, in Jones (ed.): 39-76. Jones, R. & T. Negerevich. 1985. A review of previous archaeological work, in Jones (ed.):1-16. Kamminga, J. & H. Allen. 1973. Alligator Rivers Environmental Fact-finding Study: report of the archaeological survey. Darwin: Government Printer. Kershaw, A.P. 1983. The vegetation record from northeastern Australia [+ or -]ka, in Chappell & Grindrod (ed.) 1983a: 100-101. 1985. An extended late Quaternary vegetation record from northeastern Queensland and its implications for the seasonal tropics of Australia, Proceedings of the Ecological Society of Australia 13:179-89. Kershar, A.P. & H.A. Nix. 1988. Quantitative palaeoclimatic estimates from pollen data using bioclimatic profiles of extant taxa, Journal of Biogeography 15:589-602. Latyon, R. 1991. Trends in the hunter-atherer rock art of western Europe and Australia, Proceedings of the Pre-historic Society 57:163-74. 1992. Australian rock art: a new synthesis. Cambridge: Cambridge University Press.

In press. Small tools and social change, in P. McConvell & N. Evans (ed.), Understanding ancient Australia: perspectives from archaeology and linguistics. Lewis, D. 1988. The rock paintings of Arnhem Land: social, ecological, and material culture change in the Post-Glacial period. Oxford: British Archaeological Reports. International Series 415. Meehan, B., et al. 1985. The wetlands sites, in Jones (ed.): 103-53. Nelson, D.E. et al. 1995. Radiocarbon dates for beeswax figures in the prehistoric rock art of northern Australia, Archaeometry 37:151-6. Nix, H.A. & J.D. Kalma. 1972. Climate as a dominant control in the biogeography of northern Australia and New Guinea, in D. Walker (ed.), The natural and cultural history of Torres Strait: 61-92 Canberra: Department of Biogeography and Geomorphology, RSPacS, Australian National University. Nobbs, M. &, R. Dorn. 1988a. Age determinations for rock varnish formation within petroglyphs: cation-ratio dating of 24 motifs from the Olary region, South Australia, Rock Art Research 5: 108-46. 1993. New surface exposure ages for petroglyphs from the Olary Province, South Australia, Archaeology in Oceania 28: 18-39. Pittock, A.B. &, J. Salinger. 1983. The climatic optimum and [CO.sub.2]-warmed earth: the Australian region, in Chappell &, Grindrod (ed.) 1983a:122-5. Richardson, N. 1992. Conjoin sets and stratigraphic integrity in a sandstone shelter; Kenniff Cave (Queensland, Australia), Antiquity 66:408-18. Roberts, R., R. Jones &, M.A. Smith. 1990a. Early dates at Malakunanja II, a reply to Bowdler, Australian Archaeology 31:94-7. 1990b. Stratigraphy and statistics at Malakunanja II: reply to Hiscock, Archaeology in Oceania 25:125-9. 1990c. Thermoluminescence dating of a 50,000-year-old human occupation site in northern Australia Nature 345:153-6. 1993. Optical dating at Deaf Adder Gorge, Northern Territory, indicates human occupation between 53,000 and 60,000 years ago, Australion Archaeology 37:58-9. 1994. Beyond the radiocarbon barrier in Australian prehistory, Antiquity 68:611-16. Rosenfeld, A., D. Horton &. J. Winter. 1981l. Early man in north Queensland Canberra: Department of Prehistory, RSPacS, Australia National University. Terra Australia 6. Schrire, C. 1982. The Alligator Rivers: prehistory and ecology in western Arnhem Land. Canberra: Department of Pre-history, RSPacS, Australian National University. Terra Australis 7. Smith, M.A. 1990. An early stone industry from north Australia. Unpublished paper presented at the 14th Congress of the Indo-Pacific Prehistory Association, Yogyakarta, Indonesia, 27 August. Stocker, G.C. 1971. The age of charcoal from old jungle fowl nests and vegetation change on Melville island, Search 2:28-30. Sullivan, H. 1988. Rock engravings in Kakadu National Park. Unpublished paper presented at the First AURA Congress, Darwin, 29 August. Tacon, P.S.C. 1987. Internal-external: a re-evaluation of the `x-ray' concept in Western Arnhem Land rock art, Rock Art Research 4:36-50. 1989a. Art and the essence of being: symbolic and economic aspects of fish among the peoples of western Arnhem Land, Australia, in H. Morphy (ed.), Animals into art: 236-50. London: Unwin Hyman. 1989b. From Rainbow Snakes to `X-ray' Fish: the nature of the recent rock painting tradition of western Arnhem Land, Australia. Unpublished Ph.D thesis, Australian National University. 1991. The power of stone: symbolic aspects of stone use and tool development in western Arnhem Land, Australia, Antiquity 65:192-207. 1992. Somewhere over the rainbow: an ethnographic and archaeological analysis of recent rock paintings of western Arnhem Land, Australia, in J. McDonald &, I. Haskovee (ed.), State ofthe art: regional rock all studies in Australia and Melanesia: 202-15. Melbourne: Australian Rock Art Research Association. Occasional AURA Publication No. 6. 1993a. An assessment of rock art in the Mann River region, Arnhem Land, NT. Report to the Bawinanga Aboriginal Corporation and the Djomi Museum. 1993b. Regionalism in the recent rock art of western Amhem Land, Northern Territory, Archaeology in Oceania 28:112-20. 1994a. Socializing landscapes: the long-term implications of signs, symbols and marks on the land, Archaeology in Oceania 29:117-29. 1994b. Kuninjku Kunwarrde Bim: results of the 1994 Mann River Rock Art Recording Project field season. Unpublished report to the Bawinanga Aboriginal Corporation and the Djomi Museum. Tacon, P.S.C. & C. Chippindale. 1994. Australia's ancient warriors: changing depictions of fighting in the rock art of Arnhem Land, NT, Cambridge Archaeological Journal 4:211-48. Tacon, P.S.C., M. Wilson & C. Chippindale. 1995. The semiotics of symbolic snakes: recognising Rainbow Serpents, their signs and associates in Arnhem Land rock in. Paper to be presented at the News 95-International Rock Art Congress. Torgersen, T. et al. 1988. Late Quaternary environments of the Carpentaria Basin, Australia, Palaeogeography, Palaeo-climatology and Palaeoecology 67:245-61. Wade-Marshall, D. & P. Loveday (ed.). 1988. Floodplains research, northern Australia: progress and prospects 2. Darwin: NARU, Australian National University. Watchman, A. 1987. Preliminary determinations of the age and composition of mineral salts on rock art surfaces in the Kakadu National Park, in W.R. Ambrose & J.M.J. Mummery (ed.), Archaeometry: further Australasian studies: 36-42. Canberra: Department of Prehistory, RSPacS, Australian National Unevirsity. 1990. A summary of occurrences of oxalate-rich crusts in Australia, Rock Art Research 7:44-50. Woodroffe, C.D. 1988. Changing mangrove and wetland habitats over the last 8000 years, northern Australia and southeast Asia, in Wade-Marshall & Loveday (ed.): 1-33. Woodroffe, C. D. et al. 1986. Geomorphological dynamics and the evolution ofthe South Alligator tidal river and plains, Northern Territory. Darwin: NARU, Australian National University. Mangrove Monograph 3. Woodroffe, C. D., J.M.A. Chappell & B.G. Thom. 1988. Shell middens in the context of estuarine development, South Alligator River, Northern Territory, Archaeology in Oceania 23:95-103.