The antiquity of Australian silcrete heat treatment: Lake Mungo and the Willandra Lakes

Abstract

The evolution of heat treatment for stone artefact production is a subject of major interest for our understanding of early modern humans. In this study, we examine the evidence from one region in Australia to provide a new record of the antiquity of heat treatment, explore chronological shifts in the frequency of heat treatment, and discuss the implications of these findings for early population dynamics and the technical knowledge early settlers might have brought with them. Until now, Australian heat treatment has only dated back 25000 years. This study of the Willandra Lakes, including Lake Mungo, has identified the oldest systematic evidence of heat treatment yet reported in Australia, dating to ∼42000 years. We also document time-dependent directional change in the frequency of the practice. At those early times, with over 60% of all silcrete artefacts heat-treated, we hypothesize that the practice was mastered and integrated as a recurrent technical solution to the complexities of knapping silcrete. Over time, the use of heat treatment decreased progressively until less than 10% of the artefacts were heat-treated in the terminal Holocene. This trajectory has implications for understanding the antiquity of heat treatment on the Australian continent and for investigating the factors that governed its use.

Introduction

In lithic technology, heat treatment of stone is the intentional modification of the properties of raw materials with fire. Its use for stone artefact production is a subject of major interest for our understanding of early modern humans. Heat treatment was one of the earliest processes documenting the way behavioral modern humans transformed their material world, and it may even constitute a marker for complex and modern behaviors. In Australia, heat treatment was part of the suit of technical solutions used by relatively recent populations. Until the evidence announced in this article, the oldest evidence of heat treatment in Australia had been dated back to ∼25 ka (see for example: Hanckel, 1985). However, that relatively late date created puzzles as the known antiquity of human occupation has been extended by recent work. Colonization of the continent is now known to have begun approximately 60–75 ka (Clarkson et al., 2017, Hiscock, 2017), leaving unanswered the question of whether heat treatment was employed early in Australian prehistory.

This question addresses the cultural systems of people first colonizing Australia and the character of subsequent cultural evolution. It has long been proposed that the Pleistocene occupation of Australia was associated with technology that was simple, uniform, and slowly changing (Hiscock, 2008, Hiscock, 2013). That perceived character of early technology has been partly built on the proposition that early cultural variation in groups moving from Africa displayed reductions as a result of serial founder effects (e.g. Mellars, 2006). This claim that Pleistocene technologies were simple and uniform formed the foundation for perceptions of later expansion of the complexity, diversity, and regional differentiation of lithic technology in Australia, with some researchers playing with notions of a late onset of modernity in Australia (e.g. Brumm and Moore, 2005). Such claims have been rejected by other researchers who recognize evidence of the gradual onset of technological and symbolic behaviors (e.g. Habgood and Franklin, 2008) and who recognize the low discovery probabilities of uncommon items caused by small samples and poor preservation for the earliest periods of human occupation in Australia (e.g. Langley et al., 2011). Progressively, there have been surprisingly early examples of technological diversity and public signaling revealed, suggesting that the provision of further evidence may clarify the antiquity of technological shifts in ancient Australia and test the diverse evolutionary models that have been discussed (e.g. Brumm et al., 2017, Hiscock, 2017).

As heat treatment is known in Africa at a time before the dispersal of modern humans across the old world, exploring the antiquity of heat treatment within Australia can test hypotheses about the continuities or discontinuities in traditions of lithic reduction processes. Indeed, in the context of the southern African Middle Stone Age (MSA), heat treatment has been interpreted as a proxy for several evolutionary concepts. When the antiquity of silcrete heat treatment was first discovered (perhaps even being as old as ∼164 ka, Brown et al., 2009), the only possible comparison was with flint heat treatment from the European Upper Paleolithic. Flint heat treatment is thought to require technologically complex oven-like structures that keep temperatures and heating rates low to prevent the flint from overheating (Schmidt et al., 2016). The possibility that such elaborate heating structures may have been invented early in the MSA lead some researchers to propose that silcrete heat treatment was a strong argument for an early onset of modern behaviors (see for example: Sealy, 2009), complex cognition (Wadley, 2013), and the willingness to spend large amounts of resources (Brown and Marean, 2010). While this view might not be so straightforward (i.e. early heat treatment might not have been as technologically complex as initially thought, compare: Schmidt et al., 2015, Schmidt et al., 2013b and: Wadley and Prinsloo, 2014), it nonetheless is undisputable that the MSA hunter-gatherer's decision to no longer accept the properties of the lithic resources available to them, but to deliberately transform them, was a major shift in the niche construction capacity of hominins who had long before become habituated to the production and use of lithic tools but who until this point were focused only on careful selection or sources and nodules to obtain appropriate materials.

The development of heat treatment created a new niche modification technique that facilitated more regular and widespread use of some knapping strategies and wrought permanent changes on rocks littered across the landscape, altering the availability and distribution of desirable lithic materials, and the use-character and utility of tools. Feedbacks in hominid niche construction interactions often resulted in not only immediate functional effects for human groups but also longer term modification of social organization, such as elaboration of the social scaffolding of learning and skill transmission, differentiation of roles to facilitate specialization, amplification of sharing arrangements, and so on (for a theoretical background see: Sterelny, 2012). Similar feedbacks and resulting transformations of behavioral patterns could be expected in Australia.

The antiquity of heat treatment in Australia, the effects of that technique in constructing niches within this continent, and the consequences of the new niches for broader economic and social strategies are largely unexplored. Niche construction, and the contribution of techniques in which the knapping characteristics of rock were altered, was likely to have been an important process in the early period of Australian prehistory, when humans began settling, modifying and adapting to the various environments across the continent. In this article, we start an exploration of these complex interactions by pursuing three fundamental questions: (1) when and where was heat treatment employed in Pleistocene Australia, (2) how did heat treatment solve technical problems, and (3) what niche altering effects might it have had?

To pursue those questions, we document the antiquity of heat treatment in the southeastern portion of the Australian continent and describe statistical data on its prevalence and evolution though time. To achieve this, we studied a series of sites from the Willandra Lakes system, ∼700 km inland from Australia's south-east coast. This arid region is well known for its exposures of late Pleistocene land surfaces containing lithic assemblages, such as on the lunette of Lake Mungo, which yielded archaeological materials that have contributed significant information about the early settlement of Australia.

From the late 1960s onwards, archaeological fieldwork at Lake Mungo, as well as connected lakes to the north and south, established a record of late Pleistocene settlement and land use that now extends back at least 45 ka (Bowler et al., 1970, Shawcross, 1998, Grün et al., 2000). It is this corpus of assemblages, collected more than half a century ago, that we analyze in this article. The region offers substantial opportunities for tracing technological evolution. Resolution of the archaeological record in these deflating landscapes is being investigated (see for example: Stern, 2015), and the initial typological depictions of lithic assemblage variability has been revised to reveal what may be the underlying structure of the economics of lithic production and transportation.

The primary rock used for knapping in this landscape is silcrete, and this is most commonly collected and quarried from outcrops on the western side of lakes (see for example: Williams, 1991), while the lithic-rich stratified sites are found in the lunettes on the eastern side of the lakes, implying an eastward flow of lithic materials. The lack of silcrete on many land surfaces makes the transportation of rock across these regions essential for provisioning of tools where activities took place (Webb, 1993). There were significant shifts in local land use over time, particularly relating to oscillations in lake levels which created water barriers to movement through the landscape, as well as creating redistributions of food resources. There is now evidence that at Lake Mungo foragers carried lithic artefacts across at least small water gaps, possibly implying the transportation of silcrete on water craft at high water stand 24 ka (Fitzsimmons et al., 2015), but it is likely that many silcrete artefacts on lunettes were transported to the eastern side of lakes by being carried around the lake margins, increasing provisioning costs (Hiscock and Allen, 2000). Encounter rates with silcrete sources and costs of transporting rocks across these sandy and lacustrine environments would have varied over time, partly in response to lake levels as well as with site residency time and technology, and hence, we would expect repeated reconfiguration of technology and patterns of material transport over time as the economics of foraging shifted.

Articulations of technological practices and lithic economy have principally been explored for geographical variability. For instance, Hiscock and Allen (2000) hypothesized that assemblage variability through time and space in the Willandra region likely reflected these shifting costs and suggested that the general trend toward eastwards transportation of silcrete artefacts may have been associated with a variety of technological strategies that would reduce the frequency of trips to resupply the lithic toolkit. They employed previously collected typological data sets to show that the patterns of assemblage variation were consistent with initial reduction on the western shore, selection of specimens with greater reduction potential for transport eastwards, as well as the application of somewhat different reduction treatment on retouched flakes and cores discarded on the eastern shore, perhaps to extend reduction. Testing this set of ideas by reanalyzing some assemblages from Lake Mungo largely confirmed the model, although it indicated the early stage reduction on the western shores was more extensive than had been visible in the typological data (Roy, 2013). However long-term chronological trends in lithic technology have not been systematically described in the Willandra region, and in particular the existing data has not allowed an assessment of the use and antiquity of heat treatment. Heating of rocks can enhance the control of fracture (Schmidt et al., 2019) and possibly extend utility in the kinds of coarse-grained silcrete found near the Willandra Lakes, making heat treatment one option for adjusting economic costs over time. Since the Willandra Lakes have a ∼45 ka sequence of silcrete artefact production (Shawcross, 1998), and since in many situations heat treatment is readily recognizable on silcrete, this region provides a remarkable opportunity for assessing the evolution of heat treatment in the region and considering its possible role in facilitating landscape adaptations.

Heat treatment of stone for tool knapping has been documented from all continents and prominent examples are known from North and South America (Crabtree and Butler, 1964, Wilke et al., 1991, Frank, 2011), the European and Asian Upper Paleolithic and Mesolithic (Bordes, 1969, Flenniken, 1987, Tiffagom, 1998, Eriksen, 2006), and the European Neolithic (Binder, 1984, Léa, 2005, Schmidt et al., 2013a). Most of these contexts have in common that fine-grained silica rocks like flint and chert were heated. Although the use of silcrete (a pedogenic silica rock that like flint and chert offers the possibility to significantly improve its knapping quality with heating) is known from almost all continents (see among others: Nami, 2017, Wragg Sykes et al., 2017), heat treatment of silcrete is known from only two distant regions, Africa's cape region and Australia. The first of these regions lies near the geographic origins of the earliest Homo sapiens and silcrete heat treatment was practiced there from the second part of the MSA on (Brown et al., 2009). The second region, Australia, is near one endpoint of the global expansion of H. sapiens, and new information on the early history of heat treatment on the continent will potentially address early population dynamics and the technical knowledge early Australians might have brought with them. Unfortunately, early sites that yielded sufficient silcrete artefacts, where such questions can potentially be studied, are extremely rare in Australia. While our knowledge of prehistoric heat treatment in Australia dates back to the 1970s and studies documenting Australian heat treatment are numerous (see for example Corkill, 1997, Hanckel, 1985, Hiscock, 1993, McDonald and Rich, 1994, Rowney, 1994 for the Sydney and Hunter Valley regions; Flenniken and White, 1983 for Tasmania; Cochrane et al., 2012 for central Queensland; or Akerman, 1979 for the Kimberleys), the vast majority of the known sites are either subrecent, undated (Akerman, 1979, Flenniken and White, 1983, Rowney, 1994), or of Holocene age (McDonald and Rich, 1994, Rowney, 1994, Cochrane et al., 2012). The oldest published lithic sequence documenting heat treatment in Australia comes from the eastern seaboard where sites such as Burrill Lake have the practice dated back to approximately 25 ka (Hanckel, 1985, Schmidt and Hiscock, 2019). While there are substantially older stratified sites elsewhere in Australia they are often characterized by (i) low numbers of lithic artefacts in the earliest layers and/or (ii) low numbers/proportions of silcrete artefacts. The Willandra Lakes are an exception to this pattern, with large assemblages of silcrete and long chronological sequences, and we have therefore focused on this region, conducting a large scale study on six previously collected sites. These sites provide samples of silcrete artefacts dating from ∼42 ka, the oldest being associated with Australia's oldest human remains, until only a few hundred years before the first European settlers arrived on the continent. Our study focusses on two questions:

• Do the Willandra sites allow us to document the antiquity of silcrete heat treatment in Australia, and if so can we observe the onset of heat treatment at a particular point in time? If heat treatment is absent from the oldest studied assemblages, it may be possible to estimate the time of its local invention or introduction. But if in the oldest Willandra assemblages there is evidence that heat treatment was already fully integrated into silcrete reduction, that would constitute evidence that the technique was used in the early millennia of colonization and may have been used by the first settlers.

• Are there directional trends in the evolution of heat treatment-related behaviors though time? This question is prompted by our recent findings of directional change in the region centered around Sydney on the eastern seaboard. There, a trend toward increasing numbers of heat-treated artefacts in assemblages was found (Schmidt and Hiscock, 2019). That study did not establish whether increasing use of heat treatment over time was a general trajectory active in very different environments across Australia, reflecting a continental-wide selective response, or if there were different evolutionary trends in different regions. This study therefore tests the geographic uniformity of evolutionary change in heat treatment practices in Australia.

We analyze 8 sites for this study, and together, they provide a window into a period from the late Pleistocene to almost the end of the Holocene (Table 1). Their geographical position is shown in Figure 1. We present the samples in chronological sequence, beginning with the earliest.

The 2 oldest assemblages come from the lunette on Lake Mungo. One of our Mungo samples, labeled here the “LM lunette sample”, was collected west of the WLH1 cremation site by J.G. Robinson in 1978 (Robinson, 1980) at a period when more systematic excavations were being conducted in this area (Shawcross, 1998). We did not have access to the material from the large Shawcross excavation. The age of artefacts in our LM lunette sample can only be roughly estimated because they lay on a deflated surface and hence may have derived through erosion from two strata on the Lake Mungo lunette: the ∼17–23 ka old Zanci Unit (as dated by TL and 14C, see for example: Bowler et al., 2012) and the ∼30–49 ka old Mungo Unit (OSL dated, see: Bowler et al., 2003). Artefacts from both layers formed a palimpsest sitting on the modern land surface and cannot be separated in our assemblage (Robinson, 1980). Consequently, this Lake Mungo lunette sample contains artefacts that potentially cover an age bracket of ∼17–42 ka, corresponding to the youngest age of the Zanci Unit and the age of the Mungo 1 cremation at the top of the Lower Mungo Unit (Bowler et al., 2003). We also examined a second sample from the Lake Mungo lunette, which was collected during fieldwork surrounding the 1974 excavation of the WLH3 burial. Some of these artefacts were labeled “in situ”, strengthening the hypothesis that their stratigraphic origin is the top of the Lower Mungo Unit, like the burial itself, or in the overlying Upper Mungo Unit. We therefore suggest these specimens have an antiquity of ∼30–42 ka, roughly corresponding to the upper limit of the Mungo Unit and the age of the burial itself.

All of the other assemblages we studied came from sites on the other Willandra Lakes. We additionally studied the assemblage from Top Hut 1—Hearth 4 (H4), excavated from the outer lunette of Lake Arumpo by I. McBryde in 1974 and 1975. Debate on the exact age of the Top Hut 1 assemblages was triggered by inconsistencies between ¹⁴C ages obtained from shellfish and charcoal (see for example: Allen, 1972, Bowler and Thorne, 1976, Hope, 1993). The sometimes significantly older shell dates at Top Hut have been explained by several factors (mainly taphonomic and stratigraphic in origin: some of the dated shells might not be associated with human occupations), and no clear consensus was reached by the excavator. We therefore retain the conservative charcoal date of 34,226 ± 465 CalBP (proposed by Gillespie, 2016) as being unambiguous, while noting that older shell dates (40,699 ± 1264 CalBP, also see: Gillespie, 2016) were also obtained from Top Hut 1. For the same reasons, we accept the dates of 25,296 ± 397 CalBP (emu eggshell) and 25,349 ± 383 CalBP (fish otolith) for hearth 3 (H3) at Top Hut 1 (¹⁴C dates calibrated for this study, CalPal_2007_HULU; also note that older ¹⁴C dates obtained in the 1970s and 80s are close to the method's limits then and should be regarded with caution).

We studied specimens from Leaghur Backshore II, a location collected by Harry Allen at the western end of Lake Leaghur. Associated shell gave an age estimate of 18,950 ± 289 CalBP (Allen, 1972). A similar date is available for a different assemblage, Mulurulu III, which was collected north of Lake Mulurulu. Allen (1972) published 2 dates of 18,787 ± 360 CalBP (shell) and 15,348 ± 1375 CalBP (charcoal) for this assemblage, dates which further highlight the inconsistency of dates obtained from shell and charcoal. As the Mulurulu III dating has not been discussed in more detail since Allen's work, we retain both dates as reasonable estimates of the Mulurulu III assemblage.

We studied two assemblages from the southern margin of Lake Garnpung. The Garnpung 1A + B assemblage, collected from a grey sand, was initially associated with two ¹⁴C dates of 18,618 ± 386 CalBP (shell) and 16,867 ± 765 CalBP (charcoal), whereas the Garnpung 1C assemblage was found in an adjacent consolidated brown sand dated to 3857 ± 115 CalBP (Allen, 1972). Allen (1972) initially proposed that both the 1C and 1A + B assemblages were late Holocene in age, but subsequent considerations of the published typological counts have concluded that the composition of these assemblages are a consequence of multiple factors and that they are potentially palimpsests of specimens from different time periods (Hiscock and Allen, 2000). We therefore retain the original ¹⁴C age estimates obtained by Allen, which implies that Garnpung 1A + B likely contains material that is terminal Pleistocene, combined with a later component. We take the ¹⁴C dates from Garnpung 1A + B as an indication of the possible antiquity for the older component, making it roughly contemporaneous but slightly younger than Leaghur Backshore. Both the radiometric and compositional evidence from Garnpung 1C are consistent with a late Holocene age.

The most recent assemblage in our sample is the Mungo Backshore site, which was collected from the western shore of Lake Mungo. Two ¹⁴C dates of 855 ± 55 CalBP (charcoal) and 735 ± 135 CalBP (shell) were obtained from this vicinity and place the site within the last thousand years (Allen, 1972). Silcrete outcrops and quarrying of them in similar localities at Lake Mungo have been studied by Williams (1991).

From these 8 previously collected assemblages, we studied a total of 3248 determinable artefacts that were analyzed: 1132 from Lake Mungo (n = 413 from the LM lunette sample; n = 94 from the WLH3 burial area; n = 625 from Mungo Backshore), 256 from Lake Arumpo (n = 62 from Top Hut 1, H3; n = 194 from Top Hut 1, H4), 418 from the Leaghur Backshore site at Lake Leaghur, 103 from Mulurulu III at Lake Mulurulu; 1338 from Lake Garnpung (n = 632 from Garnpung 1C; n = 703 from 1A  +  B). A selection of these artefacts is displayed in Figure 2. All the artefacts we studied were subjected to visual classification into heating proxy categories using the protocol described below. After this visual analysis, 100 artefacts were randomly picked from the LM lunette sample without any macroscopic inspection and set aside to be analyzed by quantitative surface measurements with the replica tape method (Schmidt, 2019). This sample provided an independent cross-check on our classification.