Sandridge Deserts - Biogeography, Human Ecology and Prehistory

Australia: The Land Where Time Began

The first Europeans to see this extremely arid part of Australia were the members of Sturt's 1844-45 expedition, to be followed by a number of other expeditions such as Warburton in 1875, Lindsay in 1886, Giles in 1889 and Carnegie in 1898. The surgeon on Sturt's expedition, a Mr. Browne, wondered if man had ever seen such a place. The later expeditions found that the there were small, scattered groups of people occupying the dunefields who were apparently flourishing.

Until recently it was believed the arid interior had not been occupied until about 15,000-10,000 BP (cf. Bowdler, 1977; Horton, 1981). Since then evidence has been accumulating that occupation of the major desert uplands and river systems had been occupied by some time in the the Late Pleistocene (Brown, 1987; Lampert & Hughes, 1988; Maynard, 1988; Smith, 1987; Smith et al., 1991). According Smith discussion has now moved to the question of the nature of the occupation (Hiscock, 1989; Smith, 1989) and the timing of occupation of different parts of the arid zone.

Attention has been drawn to the lack of evidence supporting the occupation of the major sandridge deserts before about 5,000 BP (Veth, 1989b) in which it was argued that the mid-Holocene would have been the time when the arid sandridge deserts were first occupied, proposing what has been called the 'barrier desert' theory by Smith. According to Smith, the barrier desert theory should be examined critically as it forms the cornerstone of a new model that has been explicitly proposed as a new framework for desert prehistory (Veth, 1989a, 1989b).

Smith's paper examines 4 aspects of the barrier desert theory:

Do the major sandridge deserts, the Great Sandy Desert, the Great Victoria Desert and the Simpson Desert form a coherent biogeographic unit?
Are the 'barrier' deserts different from anything prospective colonists had seen previously in other parts of the arid zone?
What were the social and technological prerequisites for colonisation of these regions?
What is the archaeological evidence for this theory?

Barrier desert theory

In a larger model, 'Islands of the Interior', of which the 'barrier desert' theory forms a part, Veth has argued for a post glacial re-colonisation of the arid zone from refuge areas occupied by humans during the Last Glacial Maximum (Veth, 1989b, following Smith 1988: 5-57, 293-343). Unlike earlier formulations it singles out the major sandridge deserts, suggesting they were regions where colonisation was independent of these palaeoenvironmental changes. He proposes that various socioeconomic changes in neighboring areas of the arid zone during the mid-Holocene led to the occupation of these regions.

It has been argued that the sandridge deserts, as well as lowland desert desert habitats, would probably have been initially occupied in pre-Glacial times, then abandoned at the Glacial Maximum, then recolonised by the expanding population following the end of the Glacial period (Smith, 1988), Veth argues that the colonisation in the mid-Holocene was the first time these regions have been occupied.

Veth uses the biogeographic terms barrier, refugia and corridor to emphasise the distinction between different habitats (cf Heatwole, 1987). Under the scheme proposed by Veth:

barriers are the major sandridge deserts, the Great Sandy Desert, the Great Victoria Desert and the Simpson Desert.
refugia are the Pilbara, the Kimberley, Flinders Ranges and the Central Australian Ranges.
corroidors are the intervening areas, the Tanami Desert, the Strzelecki Desert, the Gibson Desert and the Tirari Desert, The Nullarbor Plain and the Barkly Tableland.

A major problem for people trying to colonise the 'barrier' deserts would have been the association of dunefields with hummock grassland and drainage that was uncoordinated, according to the theory. The people would have been unfamiliar with the hummock grassland environment, and poor in plant food species if they lacked the technology or ecological knowledge required to harvest and process large quantities of grass seeds. Another problem would have been the spatial patterning of water sources that were shallow enough to reach without the technical knowledge needed to construct and maintain wells. According to this theory a range of developments, technological, economic and social, were necessary before they would be capable of occupying the barrier deserts on a permanent basis. The following list has been proposed (Veth, 1989b: 83):

Instruments needed to be developed for the working of desert hardwoods and wild seed processing;
detailed distribution knowledge and the methods for processing useful seed-bearing species in the hummock grasslands;
the technical ability necessary for the construction and maintenance of wells deep enough to tap the groundwater; and
the development of extended social networks.

Veth suggests these were probably developed in adjacent regions, adaptations to the local conditions under increased population pressure. According to Veth, the result was 'the emergence of regionally specific settlement and subsistence systems' (Veth, 1989b: 83). Smith suggests likening the process outlined by Veth to one of exaptation where new niches for exploitation were opened as a result of these changes, in this case, the sandridge deserts that were adjacent.

The chronological framework of the theory is based on 2 lines of argument. The first is evidence that in the Late Pleistocene groups were unable to cope with arid conditions, that is suggested by Veth to support this, citing evidence from Colless Creek Cave where it has been found that at the time of full glacial aridity stone from outside the river and gorge system was not used (cf. Hiscock, 1989). He says that as the initial occupation took place at a time before the interior became as arid as it is at the present, the first occupants of the now arid areas may not have been fully pre-adapted to a desert environment. His second line of argument is that the social, economical and technological changes that are prerequisite for a long-term move into a desert did not take place before 5,000 BP, though according to Smith he is not explicit as to the reason these are necessarily part of the package of changes. Smith suggests Veth clearly sees a need for a long adjustment period to local conditions in the corridors that are adjacent, that were re-occupied following the Last Glacial Maximum.

Sandridge deserts - a biogeographic unit

According to Smith, the first point he takes issue with is whether or not the sandridge deserts actually form a biographic unit, suggesting that Veth has implied that they do by lumping together the Great Sandy Desert, the Great Victoria Desert and the Simpson Desert, these deserts have many features in common, as well as strong contrasts with neighboring regions. Smith says the case for this is not as strong as could be supposed, as can be seen by a cursory search of the biogeographic literature.

The continent of Australia has been divided into 3 biogeographic provinces by Baldwin Spencer, the Torresian Biogeographic Province, the Bassian Biogeographic Province and the Eyrean Biogeographic Province (Spencer, 1896). since that time biogeographers have retained the basic divisions of Spencer (Archer & Fox, 1984; Burbidge, 1960; Heatwole, 1987; Horton, 1984; Johnson & Briggs, 1975; Tyler, 1990), though with disagreements over whether Cape York Peninsula and southwest Western Australia should be regarded as separate provinces. Whichever scheme is supported, the continental interior of Australia is still recognised as a single biogeographic province, though with different names given by different authors, Eyrean, Eremean or Sturtian. There has been a proposal for a preliminary subdivision of the Eremean flora into northern and southern elements (Diels, 1906, cited in Carolin, 1982) and (Tate, 1896), and similar suggestions have been made more recently (e.g. Nix, 1982). The significant fact is that biogeographers do not suggest the sandridge deserts are a distinct unit different from the remainder of the arid zone. Those biogeographic divisions that are recognised within the arid zone divide the arid zone into northern and southern zones crosscutting any grouping of the 3 sandridge deserts as a single unit.

Barrier deserts - differences

Extensive dunefields, uncoordinated drainage systems and a hummock grassland understorey are features that are common to all 3 of the major sandridge deserts, though they also have sufficient differences to suggest different opportunities in the form of plant resources, landforms and hydrology, for any humans wishing to occupy them. The dominant vegetation in the Great Sandy Desert is hummock grassland, either Triodia or plectrachne, with Eucalyptus forming an open tree of shrub layer in the northern part, and in the south, mixed desert Acacias. There are extensive tracts of mulga (Acacia aneura) woodland, mallee (Eucalyptus gongylocarpa) and Casuarina cristata woodland in the Great Victoria Desert. In the Simpson Desert the sandhill canegrass (Zygochloea) is the dominant dune vegetation in the southeastern third of the Simpson Desert, not Triodia and Plectrachne (Atlas of Australian Resources, 1990). Smith suggests differences in the distribution of important plant food species could be expected because the differences in light, thermal and moisture regimes that result from the wide latitudinal differences between the 3 deserts, 19^o S-29^o S. In the arid zone, 2 broad plant groups have been recognised that have characteristic temperature response patterns, one in the north and the other in the south, and in the coastal part of the region, a wide overlap zone (Nix, 1982: 64).

According to Smith, among the 3 deserts there are significant differences in landforms and landscapes. Of the 'barrier' deserts, the Simpson Desert forms a much smaller region than the other 2, and differs from them in some important characteristics (Graetz et al., 1982). 8 rivers that originally formed part of the Lake Eyre catchment, having been severed by the dunefield formation, the Todd, Hale, Illogwa, Plenty, Hay, Field, Mulligan and Kallakoopah, that flow deeply into the dunefield. Rivers that flood out into the Simpson Desert at the present occasionally channel large amounts of floodwater into the dunefield (Kotwicki, 1986). The dunes in the southeastern part of the Simpson Desert are formed of pale pelletal clay (Wasson, 1983) instead of the red siliceous sand comprising the bulk of the dunes in the 3 regions. There is also an extensive belt of closely spaced playas in this sector of the dunes.

Barrier deserts - new challenges?

The premise that the 'barrier' deserts represented a habitat that was fundamentally different from anything they people had ever encountered before, either in the corridors or the Glacial refugia, is the basis of the theory. Smith suggests this does not stand up to detailed scrutiny, especially when comparing 'barrier' deserts to the adjoining corridors or refugia.

The move to the 'barrier' deserts would, at least in many cases, have been to a poorer environment than that of the environment from which they had moved, though in terms of structure and composition it would have been basically familiar. As there are large areas of hummock grassland, dunefields and drainage that was uncoordinated in areas outside the 'barrier' deserts, the combination of these features would not have been new to the colonists. It has been suggested that before moving into the 'barrier' deserts the people would most likely have visited them opportunistically, especially after good rain (Veth, 1989a: 229;1989b: 81), permitting the people to accumulate specific local knowledge regarding water supplies and other resources, possibly over several generations after occupying the adjacent corridors and refugia. Smith suggests that any barrier posed by the sandridge deserts to colonisation would have been associated with their aridity, not with their biogeographic characteristics, the difference found in the sandridge would have been more in degree than kind.

Floristic affinities - other desert habitats

Bigeographers believe that the flora and fauna of the 'barrier' deserts are strongly linked to those of the refugia and corridors that are adjacent, and are not areas of ancient endemic flora and fauna (Greenslade & Halladay, 1982; Schodde, 1982). The basis of this belief is the floristic gradients that have been found from north to south across the Great Sandy Desert (Burbidge & McKenzie, 1983: 82-3) and in the Simpson Desert, from west to east (Fatchen & Barker, 1979). In the Great Sandy Desert, the northern sector has many species of vertebrate and plant in common with the Kimberley region, that is adjacent to it, leading to the suggestion it could be considered to be a formal interzone between the Torresian/tropical biogeographic province in the north and the Eryean/Eremean biogeographic province to the south.

The endemic genera of hummock grasses (aka spinifex) Triodia and Plechtrachne, are also widespread, so not restricted to the sandridge deserts, an understorey of hummock grasses covering 26.9 % of the Australian continent (Atlas of Australian Resources, 1990, Vol. 6, Table 1) though pure hummock grasslands are not common. Both formations are present in regions that have been classified as corridors or refugia and in 'barrier' deserts (Veth, 1989b: Fig. 1). In places such as the Pilbara and the Great Sandy Desert, species inhabit both upland and sandridge desert habitats (Jacobs, 1982: 288). In prolonged droughts Triodia and Plechtrachne both die (cf Beard, 1969). It has been suggested, based on a study of speciation in these 2 genera, that during the Last Glacial Maximum hummock grasses were probably confined to refugia outside sandridge deserts, recolonising the sandridge deserts after the close of the LGM (Jacobs, 1982: 290), leading to the suggestion that humans would have been familiar with these taxa in areas that Veth identifies as glacial refugia, long before they colonised the sandridge deserts.

The acacias comprising an open shrub layer in the 'barrier' deserts also display this pattern. There are a number of species, that are highly variable, such as Acacia ligulata and A. aneura, across the arid zone. A high proportion of species of Acacia in the Great Sandy Desert have been found to also be present in neighboring regions, especially the Tanami Desert (50-64 %) and 41 % with the Central Australian Ranges (Maslin & Hopper, 1982). A similar pattern has been found for the Great Victoria Desert, 54 % of Acacia species present there are also present in the Central Australian Ranges. It has been pointed out that there is a pattern in the distribution of groups of species that are closely related in which 1 taxon in a related pair is present on the periphery of the sandy deserts and the other member of the pair in present within the desert (Maslin & Hopper, 1982: 311). Data have been provided showing that within the arid zone there is widespread distribution of species comprising the sandridge flora of central Australia (Buckley, 1982).

Smith suggests that it seems highly unlikely that the flora of the sandridge deserts was unfamiliar to the people who were trying to colonise them, having to learn to cope with them. It has been pointed out that before people moved into the arid zone they would have had the opportunity to acquire at least part of the ecological knowledge that would be essential for successfully settling in central Australia while they were still in northern Australia (Golson, 1971). Any people colonising the 'barrier' deserts from refugia or corridors would have had the same opportunity. Comparing the species of plant being exploited for seed by the people of the Great Sandy Desert with those exploited in central Australia provides a demonstration of this. Of the plant species used for seeds in the Great Sandy Desert (Veth & Walsh, 1988: Appendix 1) 56 % (23 species) are also used for seeds in central Australia (Latz, 1982: Table 4; O'Connell et al., 1983). There were also 3 plant species used for seed in central Australia but in the Great Sandy Desert they were used for other purposes.

Spatial patterning - water resources

Smith presumes that when Veth stresses the significance of uncoordinated drainage he is pointing out that the poor water resources of the 'barrier' deserts were also much less patterned in spatial distribution and predictable, which is true, while playing down the significance of local drainage networks, such as the Ruddall River in the Great Sandy Desert, which in the desert landscape would give clues to where water could possibly be found in soakages, waterhole and wells.

Smith suggests that Veth assumes that the spatial distribution of water resources will be haphazard in the absence of a coordinated drainage system, overlooking the influence of underlying geology, such as palaeochannels (Graaff et al., 1977) in the structuring of water resources. Surface expression of this patterning, in the form of outcrops, such as calcrete, or the vegetation cover (Griffin, 1990). In the southern part of the Simpson Desert, the mikiri soakages is an example (Hercus & Clark, 1986), that appear to cluster along the northern margin of a playa chain marking the bed of a relict lacustrine basin (Loffler & Sullivan, 1979). The presence of the Triodia pungens alliance in hummock grasslands is another example of how buried palaeochannels and an associated shallow watertable (Arakel, 1986) can indicate on the surface its presence (Griffin, 1990: 443).

According to Smith, a hallmark of much of the arid zone, that includes some of Veth's refugia and corridors, is an uncoordinated drainage , so it is unlikely to have been unfamiliar to people moving into sandridge deserts for the first time.

Desert habitats - ranking

The dunefields are assumed by the 'barrier' desert theory to be a fundamentally more difficult environment for humans than other habitats of desert lowland landscapes such as stony desert, sandplain or karst, a distinction being made between 'barrier' deserts and corridors. According to Smith, this is not supported by more recent ecological studies that found that topographic variety increases runoff, nutrients and water being concentrated in such a way that more frequent plant growth is allowed than on flat landscapes (Noy-Meir, 1985; Shmida et al., 1986; Stafford-Smith & Morton, 1990). A dunefield would be expected to be more productive than a sandplain, all else being equal. It has been found in another context that in central Australia a greater variety of food plant species existed in spinifex habitats than in woodland or water course habitats (Latz & Griffin, 1978: Table 10), indicating that the poorest of desert habitats is not invariably hummock grasslands.

Smith asks whether the distinction between barriers and corridors in Veth's theory can be realistic, suggesting that, according to Veth's theory, the Nullarbor Plain, Gibson Desert ofrTanami sandplain are more favourable environments for humans than the Great Victoria Desert and the Great Sandy Desert, suggesting they are not in fact more favourable. Smith points out that Veth's implied habitat ranking is not supported by his subsistence pattern data for the Gibson Desert and the Great Sandy Desert. Veth has noted that edible roots and tubers, such as Vigna and Cyperus are not found in the Gibson Desert, that he has classified as a corridor, this desert having a much more limited suite of plant foods than are found in the Great Sandy Desert - 38 species in the Gibson Desert compared to 89 in the Great Sandy Desert (Veth, 1989a: 24-5). Another point made by Smith is that Veth has noted that in the Gibson Desert the subsistence round was non-seasonal and opportunistic, as a result of the lack of permanent water sources and low rainfall that was unreliable while in the Great Sandy Desert there was a predictable aggregation and dispersal cycle that depended upon a series of permanent and semi-permanent water sources, as well as reliable rain in the summer (Veth, 1987).

Palaeoenvironmental considerations

Smith suggests the possible role of palaeoenvironmental changes involved in accessibility of the sandridge deserts is glossed over by the 'barrier desert' theory, the question remaining as to whether 'barrier' deserts are susceptible to climatic changes, as well as whether in the past they have been more accessible to human occupation at any particular time. According to Smith, Veth has predicted evidence for an ebb and flow of settlement in the corridors, clearly agreeing that the human populations would have been affected by climatic fluctuations, though suggesting that specific adaptations were required before occupation of the 'barrier' deserts could take place, rather than climate amelioration.

An argument has been put forward (Ross, 1989) that the arid zone, that includes the major sandridge deserts, is an environment that is very stable, not being likely to have been transformed in the Holocene by the climatic shifts that occurred over that period of time. The implications of this are that the major vegetation formations, such as low open woodland, chenopod shrubland or hummock grassland, maintained the distribution of the present, despite climatic shifts, such as from the climate of the Early Holocene, that was optimal, to that of later times when the conditions became drier and cooler. Smith suggests that in this respect she is correct, as there are other factors besides climate that affect the vegetation of these regions, such as soil nutrient levels that are extremely low (Beard, 1969; Stafford-Smith & Morton, 1990; Winkworth, 1967). Even if rainfall was higher than at the present the low nutrient levels would have made them unfavourable for many plant species.

According to Smith, this argument was developed to show that the effect on the populating of the arid zone by changes in the environment during the Holocene was negligible (Ross, 1989; Ross et al., 1992), though Smith believes that the effects of quite small shifts of climate can potentially impact the distribution of human populations in the region, suggesting issues of scale are important. According to Smith, the distribution and availability of critical resources, especially potable water, as well as plant and animal foods, would be impacted by even small climatic shifts, though the sandridge deserts may not have been transformed. In the sandridge deserts many soakages, rockholes and wells are not self-replenishing, being fed by regional catchments and water that infiltrates into dunes, making them vulnerable to climatic shifts and changes in temperature regimes. According to Smith, water scarcity in sandridge deserts is not the only factor making them difficult for humans to settle, there is also low productivity that forces the people to adopt a residentially mobile lifestyle, taking them over large areas to meet their needs. As any changes in water resource network distribution that result from changes in rainfall distribution and amount, can affect access to quite large blocks of country (Cane, 1987; Smith, 1988). The productivity of these areas can be impacted on by rainfall and temperature regime changes the affect the production of plant foods and changes in the amount of cover for small mammals and reptiles, and grasses that depend on summer rainfall as well as fruit that are frost-sensitive, and species of tuber producing plants.

Smith suggest that it could be argued that the 'barrier' deserts may have been more accessible to humans in the Early Holocene than at present as evidence indicates that at that time temperatures increased and the summer monsoon incursions strengthened (Singh & Luly, 1991). He also suggests that under such conditions there may well have been a more reliable network of small watering points across the major sandridge deserts and increased plant food availability. In the southern sector of the arid zone there is palynological evidence from Lake Frome (Singh & Luly, 1991) that after about 4500 BP conditions were the driest since the Last Glacial Maximum. He suggests that in the Early Holocene, as well as possibly the period before the LGM, conditions in the sandridge deserts may have been different, even though at the present they are seen as barriers.

Colonisation - prerequisites

A list of factors that are the sociological and technological prerequisites proposed by Veth (Veth, 1989b: 83):

Development of hafted instruments for working desert hardwoods and seed-processing implements;
detailed knowledge of hummock grassland seed-bearing species, their distribution, seasonality and processing methods;
the technical knowledge necessary to tap groundwater by the construction and maintenance of deep wells; and
the creation and maintenance of widespread social networks.

Seed grinding

According to the 'barrier deserts' theory, at about 5,000 BP there was a shift towards more intensive seed use in the corridors, the most detailed of Veth's arguments, as a greater reliance on seeds was required. Implicit in this is the idea that the ecological and technical expertise that resulted allowed people to occupy the hummock grasslands of the 'barrier' deserts.

Smith suggests a problem with this scenario is that the inhabitants of the refugia and corridors would have already been familiar with the ecology of the 'barrier' deserts, claiming there is good reason for presuming that there would have already been basic knowledge of the distribution and seasonality of most, and possibly all, of the basic food plants in the sandridge deserts by the Late Pleistocene.

Smith suggests that with regard to the technical expertise, a fine distinction is needed between the appearance in the archaeological record of seed-grinding implements that are specialised and the acquisition of the technological ability needed to process seeds, this ability probably being available since the Late Pleistocene, as generalised grindstones have been found in asseblages from north Australian assemblages dating from the Late Pleistocene (Kamminga & Allen, 1993; Jones, 1985; Roberts et al., 1990), southwest Western Australia (Ferguson, 1981) and the Darling Basin (Allen, 1972). As has been argued elsewhere (Smith, 1986), the appearance in the archaeological record of specialised seed grinders indicates the adoption of a new subsistence mode, not the acquisition of new technology. He suggests it would be valid to argue from this perspective that a corollary of the colonisation of these regions would be seed grinding, though not valid to regard it as the proximate cause of these moves, saying that Veth's position on this point is unclear. According to Smith, if Veth is suggesting the acquisition of seed grinding technology is a factor controlling the timing of colonisation, that is, technologically deterministic, then colonisation could have occurred long before 5,000 BP, and if he takes the alternative stance, he must, according to Smith, search for other reasons for colonisation of the 'barrier' deserts in the mid-Holocene.

Smith also challenges Veth's view that the ethnographic exploitation of seeds was a feature of the 'barrier' deserts that distinguished them from other desert habitats, saying it is not convincingly shown by Veth's data that there was a greater reliance on seeds (Acacia as well as grass seeds) in the 'barrier' deserts. In the Great Sandy Desert, of the plant species traditionally used for food 48% were seed food species, and in central Australia it was about 50 %, based on the figures of Latz (1982: Table 4), or based on the figures of O'Connell et al, (1983), that were specific for the Alyawarra, 39 %. It has also been pointed out by Veth and Walsh that there was substantial use of tubers and roots by the Martujarra, where they were staple foods. Reptiles are also a food source, and the numbers and diversity of these regions are greater than in any other comparable sandridge deserts elsewhere in the world (Pianka, 1984). Smith suggests that, based on a correlation between the variability of the environment and opportunism or behavioural plasticity, a wide range of resources might be expected to be used (Yellen, 1976), and he suggests that the use of seeds is not necessarily a feature special to 'barrier' deserts, and does not imply that the colonisation of these regions was dependent on seed use, even though they may have been a integral part of the ethnohistoric subsistence patterns.

Desert wells - landscape capital

This is another factor that Veth discussed in detail. These took the form of a narrow shaft up to 7 m deep, often dug at an angle, down to the water table, that were lined with straw and mud, and very rarely, shored up roughly. Natural conduits in calcrete and limestone that were enlarged and maintained. That human occupation of the 'barrier' deserts was dependent upon the wells for at least part of the year is made clear from the ethnography (cf. Hercus & Clark, 1986). The people of these deserts also used other techniques to conserve water such as cleaning out rockholes and covering them, and in places rockholes were enlarged with fire (Kerwin & Breed, 1981).

The author suggests that there is little doubt 'that such wells were an essential part of the infrastructure of life in the sandridge deserts'. There is general agreement that it would take some time to construct a network of such wells, representing what has been termed landesque capital (Brookfield, 1986). The author suggests that the construction of these wells required detailed local knowledge, saying it was not necessary to have a long period of technological development to achieve a sophisticated technology to account for these wells.

Another part of life in desert refugia would have been digging deep soakages where water could be obtained in dry stream beds, channels often flowing beneath the sand of the bed. Smith also suggests that people in the Central Australian Ranges in the Late Pleistocene may have had the knowledge and skills to dig such deep soakages.

The point about the accumulation of landesque capital, involving wells and rockholes, as well as the development through patch burning of a mosaic of vegetation that formed the bassis of a productive part of the landscape, in terms of the animal species that could be hunted, is said by Smith to be an important one. He also suggests it may have been necessary to construct such an infrastructure in the sandridge deserts prior to a new part of the desert being occupied on a permanent basis, suggesting that no more than 2 or 3 generations would be required to colonise these deserts.

Technological and social factors

The use of hafted implements, such as tula adzes, for the working of desert hardwoods to make bowls and dishes that were used to collect seeds, is one of the factors that Veth mentions as prerequisites for colonisation of the 'barrier' deserts. Smith suggests that though tula adzes are a very efficient and distinctive implement for working the desert hardwoods (Sheridan, 1979) they should not be regarded as essentials for colonisation of the region, pointing out that serviceable wooden bowls that were suitable for carrying various items, such small babies, water or seeds could easily be made by using a coble to remove the bark and outer wood from a eucalypt, which he observed being done.

The mitigation of stress of local resource availability resulted from the importance of an extended kinship network. Smith suggests that there is no doubt that extended kinship networks are a prerequisite for the successful occupation of the sandridge desert regions, if it is accepted that they are a key factor in this occupation, as well as the entire process of colonisation, as they can maintain social and biological viability at low population densities. As small groups of people moved beyond the frontier a premium was probably placed on such networks, but Smith suggests they would have been just as important in the colonisation of the refugia and corridor regions as for the 'barrier' deserts themselves. He also suggests that these kinship networks must have been in place by the Late Pleistocene in the Central Australian Ranges to allow the survival of a human population in the region throughout the Glacial Maximum (Smith, 1989a), especially as the overall population density may have been very low, possibly as low as ethnohistoric levels in the Great Sandy Desert, saying he sees no a priori reason to claim that the genesis of such networks took place in the mid-Holocene.

Adjoining regions - demographic pressure

In his summary of the previous discussion Smith suggests Veth has not made a sufficiently strong case for either the construction of hafted woodworking implements or an increase in the importance of seed foods being prerequisites for the colonisation of the 'barrier' deserts. Smith also believes that Veth is unable to sustain the argument that either the extended kinship networks or the construction of deep wells were innovations in the mid-Holocene, leaving the proposition that demographic pressures in the corridors led to the occupation of the 'barrier' deserts. Smith suggests that though this idea is not surprising, there is still the problem of explaining why it occurred when it did, about 5,000 BP, and not at some other time, thousands of years after settlement in the corridors, as far back as 35,000 BP.

Testing the theory

Smith has evaluated the 'barrier desert' theory based on the available archaeological evidence, as well as his criticism of Veth's premises, suggesting that a rigorous test should deal with the processes specified in the theory as well as dealing with the predicated outcome, in this case, occupation of the 'barrier' deserts in the mid-Holocene. He suggests as an example the expectation that there would be evidence for population growth in adjacent areas some time between 10,000 BP and 5,000 BP, as well as evidence of transformations, both economic and social, in the corridor regions just prior to 5,000 BP. There should also be evidence of tula adzes and seed grinding implements being associated with occupation of these regions. He also suggests that evidence should be sought of other regions with similar water resources, landforms or plant resources being occupied earlier than 5,000 BP. Veth's test of the theory is considered by Smith to be perfunctory, restricting his discussion to the possible presence of other sites in the major sandridge deserts that could be dated to earlier than 5,000 BP.

Other aspects of the theory are also not well supported, as revealed by a review of the archaeological evidence, Smith suggesting that the earliest radiocarbon dates available for the 'barrier' deserts from occupation sites are likely to represent minimum ages for colonisation of these regions. The idea that seed grinding and harvesting of wild seeds were associated with the first movement of people into these regions is not well supported by the data available to Smith, the first appearance of seed grinding implements in deposits from sites in the 'barrier' deserts occurring higher in the sequence than the basal levels, and their first appearance in assemblages from the 'barrier' deserts or adjoining regions is not until about 1500 BP, thousands of years after the earliest evidence of occupation. There is also no evidence of economic or social transformations occurring in the adjoining regions prior to colonisation of the 'barrier' deserts in the mid-Holocene.

Colonisation of the barrier deserts in the mid-Holocene?

Not a lot is known of the prehistory of the sandridge deserts in spite of the results that have been published of the work done by a number of people in the field (Cane, 1984; Hercus & Clark, 1986; Hughes & Lampert, 1980; Smith, 1988; Smith & Clark, 1993; Veth, 1987, 1989a, 1989b; Williams, 1988).

The Rudall River region of the Great Sandy Desert (Veth, 1989a) is the only part of the 3 major dunefields for which detailed regional prehistory is available. Of the 5 stratified sites that were excavated all had basal occupation deposits that were of mid- to Late Holocene age that rested directly upon bedrock. A number of factors suggest that the ages for these sites should be interpreted as minimum ages of occupation of the region.

The various sites were first occupied over a wide range of ages.
There is no direct evidence suggesting whether or not the region was occupied by 5,000 BP as a result of the lack of a stratigraphic record for the period earlier than 5,000 BP.
It appears deposits began to form following lintel blockfall at Karlamilyi rockshelter and at Jalpiyari, ruling out any simple correlation of human use of the sites and the beginning of sedimentation.

The result is that the evidence does not give any indication that the full span of occupation of the region is represented.

Smith questions the selection of the Rudall River sites as investigations suitable for testing the 'barrier desert' theory, though he acknowledges they provide useful data. It was believed by Veth that the only extensive riverine complex that is entirely in a desert region of Western Australia is Rudall River (Veth, 1989a: 20-1). Compared to other regions in the Little Sandy Desert and the Great Sandy Desert it has a relatively high concentration of permanent waters (Veth, 1989b: 89). The Aboriginal inhabitants of the region are believed to have foraged in the watercourses and floodplains of the Rudall River for plant foods in preference to the dunefield, according to Veth's ethnobotanical work. The Throssel Range, the Broadhurst Range and the McKay Range, as well as other ranges, form a peninsular of range country that is seen on maps of physiographic regions to project from the Pilbara into the dunefields (see Beard, 1969: 693; Jennings & Mabbutt, 1977). Smith questions whether the combination of hydrological circumstances, landforms and plant resources of the area as similar enough to those specified in the theory.

According to Smith, there is a problem with they way the theory is tested against archaeological data by Veth, such as the substantial differences, that have been pointed out by Smith, when comparing the actual boundaries of sandridge deserts, as shown by Veth's Fig.1, and the distribution of areas combining the dunefields and hummock grassland and uncoordinated drainage, the discrepancy being regarded by Smith to be serious enough to cast doubt on the temporal distribution of sites in such areas.

The archaeology of the other sandridge deserts is not as well known, for example at the time of writing of Smith's paper in Source 1 there were no radiocarbon dates for the Great Victoria Desert. For the Simpson Desert and its margins the first occupation sites were found to cluster around a time less than the last 3,000 years, though the available data is too sparse to determine if a regional trend is represented by this pattern, and if so, the interpretation (Smith & Clark, 1993).

The record of occupation of the Puritjarra Rockshelter, in western Central Australia, that covered a long period from the Pleistocene and mid-Holocene, is suggested by Smith to be the most compelling evidence against the proposal that the first occupation of the sandridge deserts occurred in the mid-Holocene (Smith, 1987, 1988. 1989a). In the sandridge deserts Puritjarra is in an analogous situation to the Rudall River sites, and Smith suggests it probably conforms more closely to the physical criteria that were specified in the 'barrier desert' theory, and that similar points could be made regarding Puntutjarpa rockshelter, where the occupation has been found to extend as far as 10,000 BP.

Refugia and corridors - social and economic transformations

Beginning at about 1,400-600 BP there is a substantial population increase that coincides with the appearance of seed grinding implements in archaeological sites, such social and economic changes took place in the Central Australian Ranges on the periphery of the 'boundary' deserts (Smith, 1988). At such places as Walga Rock (Bordes et al., 1983) and Burkes Cave (Allen, 1972: 138-218) similar changes are found in the corridors that mirror the changes in site use as reported from the Rudall River sites (Veth, 1989a: 175). In the Central Australian Ranges there is evidence for some form of reorganisation of activities at large ceremonial sites in the last 1,000 years (Smith, 1988: 269-92). These changes are considered by Smith to be exactly the type of changes that could be expected under the 'barrier' desert theory, they are clearly too late to have contributed to the colonisation of the sandridge deserts in the mid-Holocerne, no comparable changes being found that date to the mid-Holocene.

Colonisation - tula adzes and seed grinders

Seed grinding implements and adzes would be expected to be present in the earliest levels in archaeological sites in the 'barrier' deserts if seed exploitation and the use of hafted woodworking tools were integral to colonisation of these deserts, though this has proven to not be the case, even in the Rudall River sites. Veth has reported that no formal grinding bases were recovered from the excavations, all other formal implements, such as tula and burren adzes, as well as backed pieces, were found in very low numbers in sites that are presumed to be less that 1500 years old, based on the respective depth/age curves for each site.

Of the stratified sites, the only one from which grindstones were recovered at some depth is Karlamilyi, in which fragmented grindstones and mullers have been found down to spit 14, the layer of the implements being bracketed by date of 3180 ± 70 (WK1093) and 1120 ± 50 (WK1092) (Veth, 1989: 133) and presumably dating to about 2000-1500 BP.

There is not much evidence of seed grinding implements in the Central Australian Ranges prior to about 1500 BP (Smith, 1988: 334-8), or of elements of hafted implements, such as tula adzes, prior to about 3600-3000 BP (Smith, 1988: 333). Evidence of a few earlier examples of seed grinding implements are found on the eastern margin of the semi-arid zone that date to about 3500-3000 BP (summerised in Smith 1989b: 311), as well as possibly earlier (Balme, 1991).

Corridors and refugia - population growth

The proposal that in the corridors and refugia population density increased between 10,000 BP and 5,000 BP is difficult to test because of the lack of sites in which there is an uninterrupted record of occupation, the only known such site is the Puritjarra Rockshelter that has a continuous sequence for this time period, recording the changes that took place at the site, displaying a substantial increase in use at about 7,000 BP (Smith, 1988: 129-32). Smith suggests that the use of the site at Puntutjarpa Rockshelter may also have increased at about 7,000 BP (Gould, 1977). The limited amount of evidence is suggested by Smith to not be inconsistent with an increase in the population density in these regions in the Early Holocene.

Adaptation to aridity in the Pleistocene

It has been suggested by Veth that the marked impact of Glacial aridity that affected the land use patterns in the area of Colless Creek indicates that the Aboriginal groups in the area were not fully adapted to aridity in the Late Pleistocene. Smith suggests it is more probable that what it documents is the circumstances in a particular local area, such as the amplitude and sharpness of environmental changes that occurred in the region, with their effect on the availability of alternative resources, not the supposed levels of adaptation taking place on a continental scale.

Aboriginal groups are suggested by other archaeological evidence from the Late Pleistocene to have had the capacity to adapt to a wide range of circumstances that occurred in the arid zone. It has been suggested that the inland economies that were well developed, based on macropods and emu eggs, might have been functioning in the Shark Bay area, Western Australia, by 25,000 BP and possibly earlier, and a human population appears to have been present in the Central Australian Ranges throughout the entire period of the Last Glacial Maximum, the most arid period of time the Australian continent is known to have experienced (Smith, 1989a). At the Puritjarra Rockshelter the inhabitants would have depended upon the resources of the surrounding spinifex sandhill country throughout the Late Pleistocene and the Early Holocene, while people of the Strzelecki Desert/Coopers Creek region (Smith, et al., 1991; Veth et al., 1990) would presumably have relied upon the surrounding dunefields, for at least part of the year, in the Late Pleistocene. Smith suggests an important focus of this occupation would probably have been the riverine corridors of Coopers Creek and Strzelecki Creek, he doubts the inhabitants could have relied upon these resources, as these riverine habitats were relatively poor, suggesting that even during the Late Pleistocene an integral part of the land use may have been periodic use of the surrounding dunefields.

Sources & Further reading