Australia: The Land Where Time Began

A biography of the Australian continent 

Keep River Region, Eastern Kimberley, Australia – Comparative Occupation Records

In this paper Ward considers the occupation record of the Keep River region in the eastern Kimberley, and whether archaeological records are preserved equally within regions as well as between regions. Evidence from archaeological sites, luminescence and radiocarbon dating, were used on 8 rock shelter sequences which contained occupation sequences dating from the Late Holocene (5,000-0 years BP), though occupation as far back as 18,000 BP was found from sand sheet sequences based on luminescence dating and archaeological evidence. Ward questions to what extent the representative records from the eastern Kimberley, as well as the adjacent western Kimberley, Victoria River District and Arnhem Land regions can be compared, given that such different chronologies can be produced by work in rock shelters and sand sheet excavations. Ward also argues that it remains unclear whether apparent intensification in the Holocene is actually a product of cultural change or of research and preservation, in the absence of comparative chronostratigraphy, of rock shelters in particular.

The Keep River cultural province is in the eastern Kimberley adjacent to, and possibly related to the cultural provinces of the western Kimberley and Victoria River Districts (Taҫon et al., 1999) and Arnhem Land (Lewis, 1988). In the Keep River region the relative occupation age compared to these adjacent cultural provinces remains in contention (Fullagar et al., 1996; Roberts et al., 1997; Watchman et al., 2000). Arnhem Land is considered to be the location of the earliest evidence of occupation (Mulvaney & Kamminga, 1999), though the previous age determination of about 60,000 BP (Roberts et al., 1998b) is now being questioned, with a younger age of about 45,000 BP being proposed (O’Connell & Allen, 2004). The earliest evidence of occupation in the western and eastern Kimberley is about 40,000 BP (O’Connor, 1995) and about 20,000 BP respectively (Dortch & Roberts, 1996). Ward suggests that before any comparisons are made regarding occupation between regions it is important to consider if archaeological time periods are preserved equally within regions (Waters & Khuen, 1996), as comparisons between regions may be only random.

In this paper Ward compares the record of occupation in sedimentary sequences from 5 rock shelters with the sequences from 3 sand sheets located within the Keep River region. The sand sheets link the escarpments to the rock shelters, and the research was concentrated at 5 sites known in Aboriginal tradition as Jinmium, Goorurarmum, Punipunil, Granilpi and Karlinga. Geochronological determinations that were derived from recent (Ward, 2003; Ward et al., 2005) and research that had been published previously (Fullagar et al., 1996; Head & Fullagar, 1997; Atchison, 2000), and included dating by radiocarbon, thermoluminescence (TL), and optically stimulated luminescence (OSL). According Ward full site descriptions can be found in the earlier publications, and a discussion of the sedimentary and archaeological records (Ward et al., 2004). Keep River current data have been integrated into the records of occupation in the eastern Kimberley and Arnhem Land.

Keep River archaeological records – review of archaeological excavations

Sedimentary stratigraphy of sand sheets and rock shelters are typically comprised of loose surface sands that is charcoal-enriched, which overlies sands that are slightly more compact, which in turn overlie rubble and/or a bedrock base. Among the cultural materials are flaked and stone points, charcoal, processed seeds, ochre, bone, and glass. Stone points recovered in the Kimberley region have been dated to about 3,000 BP (Attenbrow et al., 1995; Fullagar et al., 1996: 764). For the excavations at Jinmium, Granilpi and Punipunil full details of the chronometric, sedimentary and archaeological records can be found in (Fullagar et al., 1996; Atchison, 2000; Atchison et al., in prep.) and for excavations at Goorurarmum and Karlinga (Ward, 2003; Ward et al., 2004, 2005).

The Karlinga (Karl-1) rock shelter is situated at the base of a sandstone cliff, and is sheltered behind several boulders. At 27 cm depth an age of 18,400 ± 1,400 BP was found by OSL dating of sediments was obtained. As with the sediments from the Jinmium rock shelter (Roberts et al., 1999), the sands were thought to probably include ‘saturated’ quartz that was derived from slow disintegration of the overlying and surrounding bedrock  (for more details see Ward, 2003; Ward et al., 2004), which is why this estimate is not considered to represent an occupation age. This age of 18,500 BP is also inconsistent with:

1 Estimates of radiocarbon age younger than 1,000 BP obtained from charcoal samples from the same sediments,

2 with other luminescence and radiocarbon ages for excavations in rock shelters in the Keep River region,

3 and with the presence of flaked stone and Kimberley points.

The location of the Karlinga sand sheet excavation is about 500 m from the site in the Rock Shelter. At this site the sediments are not affected by contamination from the bedrock as there was no contact with the underlying rock. At 240 cm depth an age of 18 ± 6,00  BP was obtained by OSL dating, and for the surrounding sediments TL produced a similar estimate (see Ward, 2003; Ward et al, in review, which represented a minimum age for the beginning of formation of the sand sheet (Ward, 2003). Immediately above and below the 2 cobble layers the highest density of stone artefacts, which included stone points, were found dating to about 2,500 BP by OSL and 900 BP respectively.

At the Goorurarmum excavation site (Goor-2) is in an elevated rockshelter, and the adjacent sand sheet (Goor-1) is about 20 m in front of Goor-2. Within the rockshelter the sediments produced a similar age by OSL of 500 ± 140 BP and ages of 300 ± 70 BP by radiocarbon. According to Ward these recent estimates of age are younger than that indicated by the presence of stone points at the profile base, which indicates that either the points or the older sediments containing them have been moved or reworked. Dated by TL to 14,300 ± 400 BP at a depth of 220 cm, the lowermost sediments of the adjacent sand sheets are significantly older. Below these sediments is indurated (hardened) or bedrock horizon which according to Ward may represent a surface dating to the LGM. The inversion between the OSL age of 4,300 ± 100 BP at 180 cm and the TL age of 6,100 ± 100 BP at 155 cm, assuming there are no errors in the dates, indicates secondary mixing. The upper metre of the sediment sequence contains the greatest density of artefacts which includes charcoal, stone points, and bone.

Within a cluster of boulders on the northwestern side of a large sandstone outcrop is the location of the Granilpi excavation site which displays extensive rock art (Taçon et al., 1997), while Punipunil is a long narrow rock shelter within a sheltered gorge (Atchison, 2000). Early to Mid-Holocene ages for the sequences were obtained by radiocarbon dating of fruit tree seeds at both Granilpi and Punipunil rock shelter excavations (Atchison, 2000). The Jinmium (C1) rockshelter excavations are at the base of an exposed sandstone boulder, and the sand sheet is located 10 m from the rock shelter (C1/IV). OSL and radiocarbon dating of the young sediments is supported by the seed and stone artefact chronology (Atchison et al., in prep.), in spite of a disturbance or contamination (Roberts et al/. 1998b). The published chronology of the sand sheet excavation (Fullagar et al., 1996) at Jinmium, unlike the rock shelter sediments, has never been revised. A TL age of about 76 ka BP at a depth of 6m was obtained for the sand sheet sediments between Jinmium and Goorurarmum (Ward, 2003; Ward et al., 2005). Which lends support to the chronology determined by TL dating of (Fullagar et al., 1996), which produced an age of 103 ± 14 ky BP at a depth of 5 m near the Jinmium site. It was noted by Fullagar et al., (1996) that stone artefacts are present throughout the sand sheet deposit, though the initial presence of stone points, ochre and seed artefacts was dated to about 2.9 – 3.9 ky BP.

Comparison of sequences of sand sheets and rock shelters

If the 18.5 ky BP estimate for the Karlinga rock shelter, and accepting the young Holocene chronology from Jinmium, the sedimentary sequences of the Keep River region rock shelters all have an age of mid-Late Holocene (7,000 – 0 BP). Contrasting with this, the adjacent sand sheets are comprised of sediments and associated cultural sequences which are much older, providing a possible record of human occupation that extends back to 18 ky BP. Also, greater vertical accumulation of Holocene age sediments and archaeological deposits are provided by the sand sheet excavations. Ward suggests that it appears to be likely that the absence of earlier occupation within the rock shelters is due to the geomorphology of these rock shelters being insufficient for the accumulation and preservation of sediments and archaeological material in the long term compared to the adjacent sand sheets, rather than reflecting any absence of earlier human occupation.

The situation where deposits are deeper and older outside rock shelters compared to within rock shelters has been noted previously at Native Well I, Queensland, where respective ages of (6,100 BP and 11,000 BP ) (Morwood, 1981). According to Ward, in this as well as a number of other excavations where the deposit extends out past the dripline of the rock shelter, there has generally been minimal comparison of cultural deposits (e.g., Flood, 1970; Smith, 1989) and of sediments and chronology (e.g. Jones & Johnson, 1985; Allen & Barton, 1989; Morwood et al., 1995) between the rock shelter and the area outside it. The excavation records of the Keep River region indicate greater preservation of older sequences may sometimes occur in the open sites, though enclosed rock shelters may provide greater preservation potential for older sequences (Lourandos & David, 1998). In the Keep River region the rock shelters that have young ages have ages from the Late Pleistocene for cultural deposits in the sand sheet area that is immediately adjacent. There are examples of sequences dating to the Pleistocene within rock shelters, such as Narrabulgin (Ngarrabulgan) (David, 1993), Carpenter’s Gap (O’Connor, 1995, 1996) and Riwi (Balme, 2000), though the absence of a wider contextual chronostratigraphy limits interpretation of these older sequences.

Eastern Kimberley, western Kimberley and Arnhem Land – representative records of occupation

Ward suggests re-evaluation of the representative records of occupation for the eastern Kimberley, and the adjacent western Kimberley and Arnhem Land regions, given the age discrepancy between the rock shelters and adjacent sand sheets. It has been noted that there is a lack of high quality stratigraphic information and δ13C values as a guide to a material being dated. Here the extent to which the datasets are comparable is the important consideration in the regional chronological comparisons. The quality of the dating results are referred to elsewhere (for reviews see Frankel, 1990; Allen, 1994; Roberts and Jones, 2001).

Rock shelters have comprised more than 80 % of the archaeological excavation sites in the Keep River and surrounding region that have been published, and contain a freshwater or terrestrial record. Most of these sites are of Late Holocene age, with the earliest evidence of occupation dating to 4,000 – 3,500 BP. In the Victoria River District all the sites excavated have been rock shelters, all of which are dated to the Holocene. The vast majority of archaeological investigations in Arnhem Land have also been rock shelters (Taçon & Brockwell, 1995), which apparently indicates that there was an accelerated increase in the number of these beginning at the close of the Pleistocene (Morwood & Hobbs, 1995).

In the western Kimberley the Wundadjingangnari, Idayu and Goalu are the only open midden sites, the remaining 11 archaeological sites are rock shelters that contain terrestrial and estuarine records. All 3 midden sites are located in the Mitchell Plateau where the entire record of occupation is of Holocene age. For the coastal midden sites along the South Alligator River most sites are younger than 6,000 BP (Woodroffe et al., 1988). It is not clear, however, whether the greater abundance of these Mid- and Late Holocene middens indicate a change in human occupation, or the younger material has been preferentially preserved after the stabilisation of the sea level to current levels (Woodroffe et al., 1988: 101). There are no published dated midden sites, or evidence of a marine economy, in the eastern Kimberley.

The occupation sequence extends into the Pleistocene in other parts of the Western Kimberley, but it comprises a period from about 17 – 13 ky BP which has been regarded as a cultural hiatus (Veth, 1995; O’Connor et al., 1999). Evidence of a continuous cultural presence for the past 40,000 years is apparently provided at Carpenter’s Gap (McConnell & O’Connor, 1997: Balme, 2000) in spite of the hiatus above the levels from the LGM. It is still not clear whether each of the major chronological hiatuses that have been documented in each of these excavations actually represent a ‘cultural hiatus’ during which the site was not often used and there was a low level of sediment accumulation, or represent a natural hiatus during which any sediments that had been deposited in the Late Pleistocene-Early Holocene were removed by geomorphic processes (Wallis, 2001: 105).

It is apparent that in all regions of northwestern Australia, whichever was the case, most of the sites that have been excavated have been rock shelters, the majority of these being of Holocene age. Whether the records that resulted are comparable regionally depends mainly on the identification of the natural processes, e.g. physical conditions, or cultural processes, e.g. movements of populations, which Ward suggests may explain the temporal distribution patterns (e.g. Feathers, 1997; Parkington, 1989; Ward and Larcombe, 2003). In Arnhem Land, e.g., it has been documented (Woodroffe et al., 1988) how the geomorphological development, associated with rise in sea level of the South Alligator River had influenced the distribution and preservation of shell middens, with the oldest being radiocarbon dated to 6,215 ± 100 years BP. Ward suggests that it is possible that the similar temporal distribution of midden records in the western Kimberley may be accounted for by a similar geomorphological history. More generally, in the unconsolidated soil profiles that are typical of northern Australia, the limited resolution of archaeological and environmental records often result in an approach in which temporal distribution patterns are used to identify cultural processes (Parkington, 1989; Holdaway & Porch, 1995).

Before making comparisons between site types it is first necessary to normalise them to equivalent sites that are not occupied or another measure of frequency or common datum. It may only be possible to make comparisons between different sites if the associated processes which are responsible for those shared patterns can be demonstrated (see Ward & Larcombe, 2003). There are 3 different types of archaeological site which have been compared (Lourandos, 1997: 225), rock shelter, shell middens and earth mounds of southwestern Victoria and southeastern Australia for the past 12,000 years and Lourandos (1997: 225) argues for a significant increase in the use of the site and establishment since 3,500 years BP. In the absence of normalisation, however, the frequency of sites may be a reflection of the greater perseveration of younger, and probably shallower, site assemblages. That is, the trend that is shown by the different types of site can also be explained by natural processes rather than by cultural processes, and these are not necessarily the same for each site. It is argued (Lourandos, 1997: 225) that before the terminal Pleistocene when the sites first become visible, the climate was drier, with the landscapes open and semiarid, the resources were dispersed and therefore the populations of the Aboriginals were also relatively more dispersed. There would inevitably have been poorer preservation of these earlier sites. It is also argued (Lourandos, 1997: 226) that after 2,500 BP the appearance and increase in earth mounds associated with wetlands is a reflection of more intensive use of the sites. Ward suggests that it may also indicate that the survival in such environments of earth mounds might have been limited to 2,500 years because in the words of Lourandos “no equivalent sites existed in prior times”.

When considering the distribution of archaeological sites Ward suggests questioning if the same patterns were present in occupied and equivalent unoccupied sites. It may be questioned, e.g. whether the absence of sediments of Pleistocene age in rock shelters is also observed in rock shelters that contain no evidence of occupation. It may be possible in some cases to compare equivalent sites, such as bird or animal midden sites and midden sites that were man-made. According to Ward it may require closer cooperation among sedimentologists and archaeologists during the planning of excavations, and interpretative stages (e.g. Ferrand, 2001) to answer such questions.

Intensification and the Holocene record

It is generally regarded that the Middle to Late Holocene period is one in which there was increased cultural change as indicated by alterations in stone artefacts (new types), greater processing of plants, development regionally of  art styles and an increase of occupation of older sites (Hiscock, 1984). Qualitative change can be observed in the nature of plant processing (Atchison, 2000; Atchison et al., in prep) and a reduction of stone tools (Fullagar et al., 1996, Boer-Mah, 2002). It is, however, not certain whether the qualitative changes that were apparent in the number of archaeological sites and deposits in the Holocene are actually the products of 1 or more (1) cultural change, (2) research and/or (3) preferential preservation.

It is indicated by previous research in the Keep River region, in terms of cultural change, that there was an increase in the number of sites where rock art was preserved from about 4,000 BP (Watchman et al., 2000; Ouzman et al., 2000), at about 3,000 BP the introduction of stone points (Fullagar et al., 1996: 764; Atchison, 2000; Boer-Mah, 2002: 38) and archaeological evidence of the processing of fruit seeds from at least 3,500 BP (Atchison, 2000). It is also indicated by palaeoenvironmental evidence that there was significant human interference of wet and dry rainforest in the late Holocene, 5,000 – 0 BP (Head, 1996). Though there are therefore clear indications of cultural change, Ward suggests there are also natural processes which may have influenced these records. Such as, e.g., in some of the sand sheet profiles in the Keep River region with the mean grain size decreasing as a result of illuviation (deposit of illuvium) as the finer material is concentrated in deeper horizons. A similar redistribution of cultural material may be indicated by a similar pattern with depth of mean grain size and numbers of artefacts (2 – 4 mm fraction) in the same profile. A similar pattern in the distribution of artefacts and bioturbation in sandy deposits in western Illinois has also been observed (Van Nest, 2002). It has been argued (Michie, 1983: 23) that the formation of the archaeological record is wholly attributable in some sandy areas to a dynamic system of bioturbation and gravity (see also Leigh, 2001). A need to distinguish between process-related and product-related attributes of a site, and to differentiate cultural from natural processes is indicated by these results (see also Waters & Kuehn, 1996; Ward & Larcombe, 2003).

The bias towards Late Holocene timescales, in terms of research, has been indicated by a greater representation of rock shelter sites in the published records than open sites (see Smith & Sharp, 1993; Ulm in press). Similar findings come from South Africa (Parkington, 1989), where it was considered that the extreme bias that was found in the distribution of radiocarbon dates towards caves or rock shelters mainly represented patterns of cave use that was changing, and not of prehistoric settlement in general. In the case of thorough regional surveys it was indicated (Parkington, 1989: 215), that a chronological data base could be made more relevant if the set of open site assemblages is considered. Within the regional area of interest a chronographic survey of rock shelters that were unoccupied would also provide an indication of whether the patterns that were observed in the unoccupied rock shelters were representative of natural or cultural processes. The extreme weathering that is characteristic of monsoonal climates in semi-arid areas will decrease the potential for preservation of artefacts and bias a record in a sedimentary sequence in favour of younger material. An example comes from excavations in the Jinmium Rock Shelter where in the uppermost sediments seeds were noticeably weathered and below 40 cm the number declined, which indicated that this record was influenced more by factors of preservation than cultural processes (Atchison et al., in prep.). The lack of radiocarbon ages at depths greater than 150 cm in the Keep River region most likely is a reflection of in situ organic preservation resulting from changes in the level of the water table. It has also been observed at Nauwalabila (Fifield et al., 2001) that radiocarbon ages were not reliable beyond this depth, which was coincident with the appearance of pisoliths (aka a pisoid, a concentric sedimentary grain >2 cm diameter formed as a concretion) of a prior water table that was fluctuating. There may therefore be environmental factors, past or present, limiting the age-depth range of radiocarbon dates, and in some cases this may be avoided by choosing sites at higher elevations.

Whether variation in the sedimentation rate is due to natural or cultural processes, they are an important consideration when the site use intensity is being evaluated as a function of the artefact density or fauna density (Ferrand, 2001: 547). In the Keep River region, increasing rates of sedimentation on sand sheets from less than 10 cm per 1,000 years in the Pleistocene to more than 20 cm per 1,000 years in the Holocene is a major factor in the preferential preservation of Holocene records (Ward, 2003; Ward et al.,. 2004). At Nauwalabila I in Arnhem Land, a similar site, it was also noted (Hope et al., 1995) that there was a progressive increase in the accumulation of sediment, from <1 cm/1,000 years in the Late Pleistocene-Early Holocene, prior to the past 2,000 years when it ceased completely. Exposure time and increased potential for preservation may be reduced by high rates of sediment accumulation, which may explain the observed increased artefact rates of accumulation per unit time (see also Ferrand, 2001; Ward & Larcombe, 2003).

In the eastern Kimberley region the predominantly Late Holocene record of occupation may reflect real cultural changes, though the effects of research bias towards rock shelter sites and the limitations that have been imposed by preferential preservation in semi-arid sandy environments must be considered further before interpretations of human behaviour can be further considered. Between rock shelter sites and over Holocene time periods regional comparisons may be valid, though it is not known if there is a comparable record for the Late Pleistocene of occupation in the sand sheets to those preserved in the Keep River region. According to Ward these sampling and taphonomy issues are not limited to northwestern Australia, also existing in north Queensland (Ulm in press), and in South Africa (Parkington, 1989) and North America (Marshall, 2001). Therefore the re-examination of regional datasets and deciphering the relationship between absence of evidence and evidence of absence in the archaeological record that has been preserved remains important.

Ward’s conclusion

There is evidence of occupation in the sand sheets dating to the LGM in the Keep River region, though in the rock shelters only Holocene sequences are preserved. An important consideration when making comparisons within and between regions is that different geomorphic environments can produce different records of occupation, as data about land use by humans can be overlooked when the targets for analysis are restricted to select types of site. According to Ward the research and preservation bias of rock shelters of Holocene age is apparent across northern Australia, and has been the basis for many theoretical discussions concerning intensification in the Holocene, as well as abandonment and/or gaps in the record of rock shelters. Across northwestern Australia further excavation at locations at sites outside and away from rock shelters, even at sites where there is no indication on the surface of artefacts, might test these theories. Greater consideration needs to be given to processes of site formation which may have influenced the temporal distribution of archaeological sites and deposits. In many cases a multi-disciplinary approach to other possible explanations for change should be considered before cultural regional chronological patterns are considered.

Sources & Further reading

  1. Ward, I. (2004). "Comparative Records of Occupation in the Keep River Region of the Eastern Kimberley, Northwestern Australia." Australian Archaeology(59): 1-9.



Author: M. H. Monroe
Last Updated 23/04/2016
Journey Back Through Time
Experience Australia
Aboriginal Australia
National Parks
Photo Galleries
Site Map
                                                                                           Author: M.H.Monroe  Email:     Sources & Further reading