Content uploaded by Ben Marwick
Author content
All content in this area was uploaded by Ben Marwick on Oct 01, 2017
Content may be subject to copyright.
306 | NATURE | VOL 547 | 20 JULY 2017
ARTICLE doi:10.1038/nature22968
Human occupation of northern Australia
by 65,000 years ago
Chris Clarkson1, Zenobia Jacobs2,3, Ben Marwick3,4, Richard Fullagar3, Lynley Wallis5, Mike Smith6, Richard G. Roberts2,3,
Elspeth Hayes3, Kelsey Lowe1, Xavier Carah1, S. Anna Florin1, Jessica McNeil1,7, Delyth Cox1, Lee J. Arnold8, Quan Hua9,
Jillian Huntley10, Helen E. A. Brand11, Tiina Manne1, Andrew Fairbairn1, James Shulmeister12, Lindsey Lyle4, Makiah Salinas4,
Mara Page4, Kate Connell1, Gayoung Park4, Kasih Norman1, Tessa Murphy4 & Colin Pardoe13
The date when humans first arrived in Sahul—the combined
Pleistocene landmass of Australia and New Guinea—remains a con-
tested issue. Resolving this question has important implications for
debates about the timing and rate of dispersal of modern humans out of
Africa and across south Asia1–3, and when and where genetic material
was transferred between archaic hominins and modern humans
4–7
. The
impact of humans on Australia’s ecosystems has also been a long-stand-
ing topic of discussion8–12.
Current estimates of the time of initial human colonization of
Australia range from 47 thousand years ago (ka) to around 60 ka
(refs 9, 12–23). A key site in this debate is Madjedbebe, a rock shelter
in northern Australia formerly known as Malakunanja II, which was
excavated in 1973 (ref. 24) and 1989 (ref. 13). The earliest artefacts
recovered from the latter excavation included stone tools and ground
ochre pieces deposited between about 60 and 50 ka, based on thermo-
luminescence and optical (optically stimulated luminescence, OSL)
dating of the surrounding sediments13,16. These ages, the depositional
context of the artefacts and their significance have proved contentious,
mainly because of the lack of detailed description of the artefacts and
concerns over stratigraphic disturbance of the deposit and how the
artefacts relate to the dated sediments19–22,25–30. Here we report the
results of new excavations at Madjedbebe, concentrating on evidence
regarding the age and stratigraphic integrity of the deposits and asso-
ciated artefacts in the zone of initial human occupation.
Excavations and stratigraphy
Madjedbebe rock shelter is located on the western edge of the Arnhem
Land plateau in the Northern Territory, in Mirarr Country (12°30′ S,
132°53′ E, approximately 20 m above sea level; Fig. 1a, b and Extended
Data Fig. 1b, c). The 1989 excavation13 revealed cultural deposits
starting at 2.6 m depth below the surface, with a peak in artefact density
at 2.5–2.3 m depth below surface. Silcrete flakes, ground ochre, a grind-
stone and more than 1,500 stone artefacts were recovered from the
lowest occupation levels22. The associated sediments were dated to
around 60–50 ka using thermoluminescence methods, with total uncer-
tainties of 16–20 thousand years (kyr) at the 95.4% confidence level
15, 25
.
These ages made Madjedbebe the oldest human occupation site
known in Australia; two of these samples were subsequently dated by
single-grain OSL methods, which were then under development, and
these ages supported the early thermoluminescence chronology
16,29,30
.
We conducted new excavations at Madjedbebe in 2012 and 2015 to
obtain additional artefacts and sediment samples for high-resolution
OSL dating from the zone of the initial occupation, in particular.
We excavated twenty 1 × 1 m squares adjacent to, and enclosing,
the original excavations (Fig. 1c) to a maximum depth of 3.4 m. The
three-dimensional coordinates of approximately 11,000 artefacts and
other anthropogenic features (hearths, burials and pits) were recorded
and samples were collected for chronological, geoarchaeological and
macrobotanical analyses. We focus here on the northwest squares of the
excavation (southwest faces of B4–B6 and northwest face of C4), where
the frequency of artefacts and the number of OSL samples is greatest
(Extended Data Fig. 1a) and refer to other squares for ancillary data.
The basal deposits consist of culturally sterile orange sands. The
lowest artefacts were recovered from the overlying unit (around 0.7 m
thick and composed of well-sorted medium–coarse pink sand), which
dips at a low angle (around 5°) towards the front of the shelter. The
upper boundary of this unit occurs at 2.0–2.5 m depth (increasing
with distance from the back wall; Extended Data Fig. 1a), where it
The time of arrival of people in Australia is an unresolved question. It is relevant to debates about when modern humans
first dispersed out of Africa and when their descendants incorporated genetic material from Neanderthals, Denisovans
and possibly other hominins. Humans have also been implicated in the extinction of Australia’s megafauna. Here we
report the results of new excavations conducted at Madjedbebe, a rock shelter in northern Australia. Artefacts in primary
depositional context are concentrated in three dense bands, with the stratigraphic integrity of the deposit demonstrated
by artefact refits and by optical dating and other analyses of the sediments. Human occupation began around 65,000
years ago, with a distinctive stone tool assemblage including grinding stones, ground ochres, reflective additives and
ground-edge hatchet heads. This evidence sets a new minimum age for the arrival of humans in Australia, the dispersal
of modern humans out of Africa, and the subsequent interactions of modern humans with Neanderthals and Denisovans.
1School of Social Science, University of Queensland, Brisbane, Queensland 4072, Australia. 2Australian Research Council (ARC) Centre of Excellence for Australian Biodiversity and Heritage,
University of Wollongong, Wollongong, New South Wales 2522, Australia. 3Centre for Archaeological Science, School of Earth and Environmental Sciences, University of Wollongong, Wollongong,
New South Wales 2522, Australia. 4Department of Anthropology, University of Washington, Seattle, Washington 98195, USA. 5Nulungu Research Institute, University of Notre Dame, Broome,
Western Australia 6725, Australia. 6Centre for Historical Research, National Museum of Australia, Canberra, Australian Capital Territory 2601, Australia. 7Department of Anthropology, Harvard
University, Cambridge, Massachusetts 02143, USA. 8School of Physical Sciences, the Environment Institute and the Institute for Photonics and Advanced Sensing, University of Adelaide, Adelaide,
South Australia 5005, Australia. 9Australian Nuclear Science and Technology Organisation, Lucas Heights, New South Wales 2234, Australia. 10Place, Evolution, Rock Art, Heritage Unit, School
of Humanities, Griffith University, Nathan, Queensland 4222, Australia. 11Australian Synchrotron, 800 Blackburn Road, Clayton, Victoria 3168, Australia. 12School of Earth and Environmental
Sciences, University of Queensland, Brisbane, Queensland 4072, Australia. 13Archaeology and Natural History, School of Culture, History and Language, The Australian National University,
Canberra, Australian Capital Territory 2601, Australia.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
article reSearcH
20 JULY 2017 | VOL 547 | NATURE | 307
grades diffusely into poorly sorted medium–coarse (light) brown
sands. Organic inclusions are rare, but spalled fragments of bedrock
are common, especially near the dripline in square B6. No stone lines,
pavements or imbricated structures were encountered during the exca-
vation. The brown sands are compacted below about 1.5 m depth, but
become softer and encrusted with carbonates closer to its diffuse con-
tact with the overlying midden (approximately 0.5 m thick). The latter
consists of brown silty sand, abundant gastropod shells, numerous bone
specimens and some plant roots, and is buried beneath a loose surface
layer of dark sandy silt containing abundant charcoal fragments. Most
archaeofaunal remains were recovered from this Holocene midden,
with some degraded bone found to a depth of 1.76 m. Midden bone
is exceptionally well-preserved, including a maxillary fragment of a
thylacine (Thylacinus cynocephalus) coated in red pigment (Fig. 2o–q).
Artefacts and depositional integrity
Artefacts occur in three dense bands (Extended Data Figs 1, 2), with
fewer artefacts in the intervening deposits. Each band corresponds to
a change in raw material use and stone working technology (Extended
Data Fig. 2a, b) and there is no size sorting of artefacts with depth
(posterior r
2
distribution, 95% credible interval: 0.0004–0.0049). These
observations imply overall stratigraphic integrity of the deposit.
The lowest dense band (phase 2) at 2.60–2.15 m depth in squares
B4–B6 represents the zone of first occupation; it contains an in situ
hearth and an assemblage of distinctive stone artefacts made mostly
from quartzite, silcrete, mudstone and dolerite (Fig. 2 and Extended
Data Fig. 2b). The assemblage includes a number of distinctive artefact
types, such as thinning flakes and snapped points (Fig. 2c, d), faceted
discoidal cores (Fig. 2h), grinding stones (Fig. 2e, f and Extended Data
Fig. 3), whole and fragments of edge-ground hatchets (Fig. 2a, b and
Extended Data Fig. 4), ground ochres (Fig. 2g) and fragments of sheet
mica (Fig. 2k)—several of which were wrapped around a large piece of
ground yellow ochre (Fig. 2k–m).
Artefact residues and macrofossil remains demonstrate exploitation
of fuel wood and a range of plant foods (seeds, tubers and Pandanus sp.
nuts; Extended Data Fig. 5b–d) from the local eucalyptus and monsoon
vine thicket forest. The middle dense band (phase 4) at 1.55–0.95 m
depth shows an increase in the use of bipolar technology (Extended
Data Fig. 2c); quartzite is rare and quartz is abundant (Extended Data
Fig. 2b). The upper dense band (0.70–0.35 m depth) is dominated
by quartz and chert artefacts with single and multiplatform cores
(Extended Data Fig. 2b, c).
Three lines of evidence suggest that post-depositional vertical mixing
of the deposit and artefact movement is restricted to depths of approx-
imately 10 cm. First, we refitted silcrete artefacts from the lower and
middle dense bands, and found 14 and 3 refits within these two bands,
respectively, but not between them (Extended Data Fig. 6); the median
vertical distance between refitted pieces is 10.6 cm. The limited down-
ward movement of artefacts is consistent with the outcome of modern
trampling experiments at Madjedbebe31. Second, burnt artefacts are
more abundant in these two bands—indicating a probable association
with intense anthropogenic burning—than in the intervening deposits
(Extended Data Fig. 7b); their vertical separation, and the presence of
intact hearths, argues against mixing and stratigraphic disturbance over
010205km
550 m
–125 m
550 m
0 m
0
0
0m
0 m
0
0m
0
0
(Malakunanja II)
Madjedbebe
Ngarradj Warde Djobkeng
Paribari Malangangerr
Nawamoyn
Nangalawurr
Nawalanja Anbangbang
Nauwalabila
Madjedbebe
Modern coastline
0 500 1,000250 km
Marine transgression
Liang Bua
Leang Burung
Maros-Pangkep
Jerimalai
Bobongara
Kosipe
Lake Carpentaria
Arnhem Land
escarpment
Bonaparte
Gulf
Flores
89
1 m
N
Datum
Grinding
Hollows
5 m
Rock
Tree
Rockwall
Prau
Gun
1
2
3
4
EDCB
–0.4
–0.6
–0.8
–1.0
–0.7
–0.6
Wheel
Contour
5
6
a
bc
73
N
Figure 1 | Site location and stratigraphy. a, Regional map showing
the location of Madjedbebe in relation to the coastline at 65 ka and the
current coastline (white line), and other ancient archaeological sites in
Australia and southeast Asia (bathymetric data GEBCO 2014 Grid, version
20150318, http://www.gebco.net). b, Location of Madjedbebe in relation to
other key archaeological sites in Kakadu National Park, and the location of
the high sea-level stand during the last marine transgression about 6–7 ka
(topographic data Geoscience Australia 1 arc-second DEM). c, Site plan
showing the 1973, 1989, 2012 and 2015 excavation squares. Squares E1–B2
were located beneath the sloping back wall.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
article
reSearcH
308 | NATURE | VOL 547 | 20 JULY 2017
depths of several decimetres. Third, micromorphological observations
indicate only small-scale reworking of the Pleistocene deposits: sand-
sized quartz grains have cappings and linked-cappings of fine silt grains
(Extended Data Fig. 7f, k, l), which represent episodes of wetting and
drying on stable surfaces that have been disturbed subsequently, but
microfauna galleries are absent.
Numerical chronology
We dated the deposits using radiocarbon (
14
C) and single-grain OSL
techniques. Most of the charcoal samples are isolated fragments, which
decrease in abundance with depth, but nine samples were collected
from in situ hearths in the northwest squares (Extended Data Fig. 5a).
Charcoal was pretreated using acid–base–acid or acid–base
wet- oxidation procedures (the latter preferred for charcoal older than
around 20 kyr
32,33
and the
14
C content measured by accelerator mass
spectrometry (see Methods). The 14C ages of 22 charcoal samples
(15 from the northwest squares) increase progressively to around
34 calibrated kyr before present () at a depth of approximately
1.6 m, but some isolated fragments have stratigraphically inconsistent
ages (Extended Data Fig. 8g). We attribute these anomalies to small-
scale mixing of the deposits through post-depositional movement of
charcoal fragments and digging of hearth pits.
OSL dating gives an estimate of the time since mineral grains were
last exposed to sunlight34. We applied this method to individual
grains of quartz35,36 from 56 samples (44 from the northwest squares),
including the measurement of four samples in two separate laboratories
bc
eh
i
d
kj
10 cm
5 cm
f
1 mm
10 cm
a
l
m
g
opq
n
Figure 2 | Artefacts from the 2012 and 2015 excavations. All artefacts
are from phase 2 except i (phase 3), f and j (phase 4). Scale bars are 1 cm
unless indicated otherwise. a, Ground hatchet head number 9 from B1/36
(phase 2). Scale bar, 5 cm. Top insets and micrographs show striations and
grinding (left; scale bar, 2 mm) and edge rounding and polish from use
(right; scale bar, 0.2 mm). Bottom micrograph and inset show polish (scale
bar, 0.2 mm) from movement inside the haft. b, Edge-ground margin on
flake UPAF51 C2/52 (phase 2). Scale bar, 5 mm. Bottom-right inset (scale
bar, 2 mm) shows striations (arrows) from use and grinding. Top-left, the
ground edge is shown viewed from the side. Top-right, the ground edge
is shown viewed from the front. c, Invasively retouched silcrete point
from C6/61. d, Silcrete thinning flake B4/43 (1989). e, Sandstone grinding
stone GS79 from B6/54. f, Mortar GS32 from B6/31, used to pound
hard plant material and with possible outline motif in the bottom-right
corner. g, Ground ochre ‘crayon’ B6/52. h, Faceted discoidal core from
C6/42. i, Conjoining ochre-covered slab (ART9) from D2/33; inset shows
fragment of mica embedded in a thick coating of ochre, with blue circles
at the < 8.5-mm-diameter pXRF sampling locations. j, Charcoal lines
and dots on sandstone piece from C2/26. k–m, Pieces of sheet mica from
C5/56 found wrapped around a large, ground yellow ochre ‘crayon’ (n).
o–q, Photographs of a maxillary fragment of thylacine or Tasmanian
tiger (Thylacinus cynocephalus) from C2/9, coated in red pigment.
o, Archaeological specimen (left) is shown relative to a modern thylacine
cast. p and q, Detail of ochred surface at 6.7× magnification (scale bar,
10 mm) and 45× magnification (scale bar, 1 mm), respectively. The
probable age of the thylacine specimen is 2.7–3.9 calibrated kyr , as
indicated by 14C ages for spit 9 in the surrounding squares (D3 and C4).
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
article reSearcH
20 JULY 2017 | VOL 547 | NATURE | 309
(Z.J. and L.J.A.) and four samples collected by R.G.R. in 1989 (KTL158,
162, 164 and 165)13. Many of the equivalent-dose distributions include
some grains with smaller values than those of the majority of grains
(Extended Data Fig. 9), which we interpret as evidence of small-scale
disturbance of the deposit; some of the OSL samples were collected in
5-cm-diameter tubes, which will also result in some time-averaging
(approximately 1,250 years at an average sedimentation rate of about
4 cmkyr−1; see Supplementary Information). The OSL ages show a
general pattern of increasing age with depth (Fig. 3 and Extended Data
Fig. 8c–f), with consistent estimates obtained for both the replicate
and 1989 samples (Extended Data Fig. 10). The ages are also in good
agreement with the
14
C chronology (Extended Data Fig. 8g). Both chro-
nologies support previous inferences16,22,30, and the additional lines of
evidence presented above, for limited post-depositional disturbance
of the Pleistocene deposits and vertical displacement of associated
artefacts.
We developed a Bayesian model based on the OSL chronology
(Fig. 3) to estimate the start and end ages for the three dense bands of
artefacts. The lowest dense band—the zone of first occupation—has
modelled mean start and end ages of 65.0 ± (3.7, 5.7) and 52.7 ± (2.4,
4.3) kyr, respectively; the first and second error terms are the modelled
age uncertainties at 95.4% probability, excluding and including the total
systematic error, respectively (see Supplementary Information). These
ages give a mean sediment accumulation rate of 4.1 ± 0.8 cmkyr−1 for
the lowest dense band. The middle dense band has modelled start and
end ages of 26.7 ± (2.2, 2.8) and 13.2 ± (1.0, 1.3) kyr, corresponding
to a mean accumulation rate of 4.4 ± 0.4 cmkyr−1, and the upper
dense band has a modelled start age of 7.1 ± (1.0, 1.1) kyr. This new
chronology confirms the stratigraphic integrity of Australia’s oldest
known archaeological site and extends the timing of first occupation
to around 65 ka, with more precise ages than those that have been
obtained previously13,15,16; the total age uncertainties are only 3–4kyr
(68.2% confidence interval) for the OSL samples associated with the
lowest dense band of artefacts.
Discussion and implications
The new excavations have yielded a much larger and more diverse arte-
fact assemblage than those reported previously
13,22
, with more than
10,000 artefacts recovered in situ from the zone of first occupation. The
improved chronological resolution for the site allows firmer conclu-
sions to be drawn about the global significance of the earliest artefacts.
The first occupants used elaborate lithic technology, ochre ‘crayons’ and
other pigments—including one of the oldest known examples in the
world of the use of reflective (micaceous) pigment (Fig. 2i, k–m). They
also collected and processed plant foods, as revealed by macrofossils
and artefact residues. Artefacts in the lowest dense band show traces
of Australia’s earliest evidence of seed grinding and pigment process-
ing, together with the world’s oldest known edge-ground hatchets37,38
(Fig. 2a).
The settlement of Madjedbebe around 65 ka (conservatively 59.3 ka,
calculated as 65.0 ka minus the age uncertainty of 5.7kyr at 95.4% prob-
ability) sets a new minimum age for the human colonization of Australia
and the dispersal of modern humans out of Africa and across south
Asia. The final stages of this journey took place at a time of lower sea
level, when northern Australia was cooler and wetter. Our chronology
places people in Australia more than 20kyr before continent-wide
extinction of the megafauna9–11 and supports an age of more than
60kyr for the incorporation of Neanderthal and Denisovan DNA into
the modern human genome1–7. It also extends the period of overlap of
modern humans and Homo floresiensis in eastern Indonesia to at least
15kyr (ref. 39) and, potentially, with other archaic hominins—such as
Homo erectus40—in southeast Asia and Australasia.
Online Content Methods, along with any additional Extended Data display items and
Source Data, are available in the online version of the paper; references unique to
these sections appear only in the online paper.
Received 30 November 2016; accepted 19 May 2017.
1. Mallick, S. et al. The Simons Genome Diversity Project: 300 genomes from 142
diverse populations. Nature 538, 201–206 (2016).
2. Malaspinas, A.-S. et al. A genomic history of Aboriginal Australia. Nature 538,
207–214 (2016).
3. Pagani, L. et al. Genomic analyses inform on migration events during the
peopling of Eurasia. Nature 538, 238–242 (2016).
4. Reich, D. et al. Genetic history of an archaic hominin group from Denisova
Cave in Siberia. Nature 468, 1053–1060 (2010).
5. Sankararaman, S., Patterson, N., Li, H., Pääbo, S. & Reich, D. The date of
interbreeding between Neandertals and modern humans. PLoS Genet. 8,
e1002947 (2012).
NW6
SW5A (UA)
SW5A (UOW)
SW5B
NW7
SW6A
020,00040,00060,00080,000100,000120,000
Age estimate (years)
6
5
A
(
UA
)
5
A
(
U
O
W
)
5B
7
6A
P7 end (3,410–0)
NW1
NW2
SW1B
NW3
Phases 6 and 7: 70 to 0 cm depth
SW2A
SW2B
NW4
SW3A
SW3B
Phase 5: 95 to 70 cm depth
NW5
SW4A
SW4B
Phase 4: 155 to 95 cm depth
SW7A (UOW)
SW7B
NW9
SW8A
NW8B
NW10
SW6B
NW8
SW7A (UA)
NE1B
SW9A
NW15
Phase 3: 215 to 155 cm depth
NW9B
SW10A
NW11
KTL164 (R.G.R.)
KTL164 (UOW)
SW11A (UA)
SW11A (UOW)
NW12
NW13
SW2C
KTL158 (UOW)
SW13A (UA)
SW13A (UOW)
KTL162 (R.G.R.)
KTL162 (UOW)
NW14
SW3C
SW4C
Phase 2: 260 to 215 cm depth
SW5C
SW8C
SW6C
SW14A
SW7C
Phase 1: 290 to 260 cm depth
P6 start (8,180–6,090)
P5 end (9,020–7,080)
P5 start (10,530–8,850)
P4 end (14,210–12,210)
P4 start (28,920–24,560)
P3 end (30,110–26,000)
P3 start (53,980–49,160)
P2 end (55,090–50,380)
P2 start (68,690–61,260)
P1 end (76,600–65,440)
P1 start (87,410–72,960)
Figure 3 | Bayesian model of the single-grain OSL ages. Ages have been
modelled in OxCal version 4.2. Only random errors are included. Pale
probability distributions represent the unmodelled ages (likelihoods)
and dark grey distributions represent the modelled ages (posterior
probabilities) obtained in this study at the University of Wollongong
(UOW). Blue distributions represent the ages obtained at the University of
Adelaide (UA) for the four replicate samples in this study, while those in
red (labelled R.G.R.) represent the two single-grain OSL ages reported in
ref. 16. The two brackets beneath the distributions represent the 68.2% and
95.4% probability ranges. Start and end boundary ages have been modelled
for each of the phases, with the age ranges (95.4% confidence interval,
random-only errors) given in years and rounded off to the closest decade.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
article
reSearcH
310 | NATURE | VOL 547 | 20 JULY 2017
6. Fu, Q. et al. Genome sequence of a 45,000-year-old modern human from
western Siberia. Nature 514, 445–449 (2014).
7. Kuhlwilm, M. et al. Ancient gene ow from early modern humans into Eastern
Neanderthals. Nature 530, 429–433 (2016).
8. Bird, M. I. et al. Humans, megafauna and environmental change in tropical
Australia. J. Quat. Sci 28, 439–452 (2013).
9. Saltré, F. et al. Climate change not to blame for late Quaternary megafauna
extinctions in Australia. Nat. Commun. 7, 10511 (2016).
10. Johnson, C. N. et al. What caused extinction of the Pleistocene megafauna of
Sahul? Proc. R. Soc. B 283, 20152399 (2016).
11. van der Kaars, S. et al. Humans rather than climate the primary cause of
Pleistocene megafaunal extinction in Australia. Nat. Commun. 8, 14142
(2017).
12. Hamm, G. et al. Cultural innovation and megafauna interaction in the early
settlement of arid Australia. Nature 539, 280–283 (2016).
13. Roberts, R. G., Jones, R. & Smith, M. A. Thermoluminescence dating of a
50,000-year-old human occupation site in northern Australia. Nature 345,
153–156 (1990).
14. Roberts, R. G. et al. The human colonisation of Australia: optical dates of
53,000 and 60,000 years bracket human arrival at Deaf Adder Gorge,
Northern Territory. Quat. Sci. Rev. 13, 575–583 (1994).
15. Roberts, R. G. & Jones, R. Luminescence dating of sediments: new light on the
human colonisation of Australia. Aust. Aborig. Stud. 1994, 2–17 (1994).
16. Roberts, R. G. et al. Single-aliquot and single-grain optical dating conrm
thermoluminescence age estimates at Malakunanja II rockshelter in northern
Australia. Anc. TL 16, 19–24 (1998).
17. Turney, C. S. M. et al. Early human occupation at Devil’s Lair, southwestern
Australia 50,000 years ago. Quat. Res. 55, 3–13 (2001).
18. Bowler, J. M. et al. New ages for human occupation and climatic change at
Lake Mungo, Australia. Nature 421, 837–840 (2003).
19. O’Connell, J. F. & Allen, J. Dating the colonization of Sahul (Pleistocene
Australia–New Guinea): a review of recent research. J. Archaeol. Sci. 31,
835–853 (2004).
20. Allen, J. & O’Connell, J. F. Both half right: updating the evidence for dating rst
human arrivals in Sahul. Aust. Archaeol. 79, 86–108 (2014).
21. O’Connell, J. F. & Allen, J. The process, biotic impact, and global implications of
the human colonization of Sahul about 47,000 years ago. J. Archaeol. Sci. 56,
73–84 (2015).
22. Clarkson, C. et al. The archaeology, chronology and stratigraphy of Madjedbebe
(Malakunanja II): a site in northern Australia with early occupation. J. Hum.
Evol. 83, 46–64 (2015).
23. Veth, P. et al. Early human occupation of a maritime desert, Barrow Island,
north-west Australia. Quat. Sci. Rev. 168, 19–29 (2017).
24. Kamminga, J. & Allen, H. Alligator Rivers Environmental Fact Finding Study:
Report of the Archaeological Survey (Australian Government, Canberra,
1973).
25. Roberts, R. G., Jones, R. & Smith, M. A. Stratigraphy and statistics at
Malakunanja II: reply to Hiscock. Archaeol. Ocean. 25, 125–129 (1990).
26. Hiscock, P. How old are the artefacts at Malakunanja II? Archaeol. Ocean. 25,
122–124 (1990).
27. Bowdler, S. 50,000 year-old site in Australia—is it really that old? Aust.
Archaeol. 31, 93 (1990).
28. O’Connell, J. F. & Allen, J. When did humans rst arrive in greater Australia and
why is it important to know? Evol. Anthropol. 6, 132–146 (1998).
29. Jones, R. Dating the human colonization of Australia: radiocarbon and
luminescence revolutions. Proc. Br. Acad. 99, 37–65 (1999).
30. Roberts, R. G. & Jones, R. in Humanity from African Naissance to Coming
Millennia: Colloquia in Human Biology and Palaeoanthropology (eds Tobias, P. V.,
Raath, M. A., Moggi-Cecchi, J. & Doyle, G. A.) 239–248 (Firenze Univ. Press &
Witwatersrand Univ. Press, 2001).
31. Marwick, B., Hayes, E., Clarkson, C. & Fullagar, R. Movement of lithics by
trampling: an experiment in the Madjedbebe sediments, northern Australia.
J. Archaeol. Sci. 79, 73–85 (2017).
32. Bird, M. I. et al. Radiocarbon dating of “old” charcoal using a wet oxidation,
stepped-combustion procedure. Radiocarbon 41, 127–140 (1999).
33. Bird, M. I. et al. The eciency of charcoal decontamination for radiocarbon
dating by three pre-treatments — ABOX, ABA and hypy. Quat. Geochronol. 22,
25–32 (2014).
34. Huntley, D. J., Godfrey-Smith, D. I. & Thewalt, M. L. W. Optical dating of
sediments. Nature 313, 105–107 (1985).
35. Jacobs, Z. & Roberts, R. G. Advances in optically stimulated luminescence
dating of individual grains of quartz from archeological deposits. Evol.
Anthropol. 16, 210–223 (2007).
36. Roberts, R. G. et al. Optical dating in archaeology: thirty years in retrospect and
grand challenges for the future. J. Archaeol. Sci. 56, 41–60 (2015).
37. Geneste, J.-M. et al. Earliest evidence for ground-edge axes: 35,400± 410 cal
BP from Jawoyn Country, Arnhem Land. Aust. Archaeol. 71, 66–69 (2010).
38. Hiscock, P., O’Connor, S., Balme, J. & Maloney, T. World’s earliest ground-edge
axe production coincides with human colonisation of Australia. Aust. Archaeol.
82, 2–11 (2016).
39. Sutikna, T. et al. Revised stratigraphy and chronology for Homo oresiensis at
Liang Bua in Indonesia. Nature 532, 366–369 (2016).
40. Mondal, M. et al. Genomic analysis of Andamanese provides insights into
ancient human migration into Asia and adaptation. Nat. Genet. 48,
1066–1070 (2016).
Supplementary Information is available in the online version of the paper.
Acknowledgements The authors are grateful to the custodians of Madjedbebe,
the Mirarr Senior Traditional Owners (Y. Margarula and M. Nango) and our
research partners (Gundjeihmi Aboriginal Corporation) for permission to
carry out this research and publish this paper. We are also grateful to
J. O’Brien and D. Vadiveloo for assistance in the field. This research was funded
through Australian Research Council grants and fellowships to C.C., B.M.,
L.W., R.F., M.Sm. (DP110102864), B.M. (FT140100101), Z.J. (DP1092843,
FT150100138), R.G.R. (FL130100116), T.Ma. (DE150101597) and L.J.A.
(FT130100195), and through Australian Postgraduate Awards to X.C., E.H.,
S.A.F. and K.L. B.M. was also supported by a DAAD Fellowship (A/14/01370),
a UW-UQ Trans-Pacific Fellowship, and UW Royalty Research Fellowship
(65-4630). S.A.F. was also supported by an AINSE Postgraduate Research
Award (11877) and a Wenner Gren Dissertation Fieldwork Grant (Gr.9260).
Radiocarbon analyses were partly funded by Australian Institute of Nuclear
Science and Engineering grants 13/003 and 15/001 to C.C., X.C., S.A.F. and
K.N. We acknowledge financial support from the Australian Government’s
National Collaborative Research Infrastructure Strategy (NCRIS) for the Centre
for Accelerator Science at the Australian Nuclear Science and Technology
Organisation. A L’Oréal Australia For Women in Science Fellowship to Z.J.
supported the re-dating of the original sediment samples. Part of this work was
undertaken on the powder diffraction beamline at the Australian Synchrotron.
We thank E. Grey, R. MacPhail, S. Mentzer, C. Miller, M. Svob, and X. Villagran for
assistance with geoarchaeological analysis, T. Lachlan and Y. Jafari for help with
OSL dating and related illustrations, and C. Matheson and J. Field for assistance
with residue analysis.
Author Contributions C.C., B.M., R.F., L.W. and M.Sm. obtained funding and
conducted the excavation. Z.J. performed the OSL dating and Bayesian
modelling. L.J.A. conducted the blind OSL dating study. Q.H. conducted 14C
dating. C.C. and B.M. analysed the stone artefacts. J.M. performed the refitting.
B.M. and K.L. conducted geoarchaeological investigations. T.Ma. performed
vertebrate faunal identification. D.C. analysed the ground ochre assemblage.
R.F. and E.H. analysed the stone artefact usewear and residues. S.A.F., X.C. and
A.F. analysed the archaeobotanical assemblage. K.C. performed microscopy
on ART9 mica. K.N. made the map in Fig. 1 and performed analysis of marine
transgression for the study region. J.H. conducted the pXRF analyses. J.H. and
H.E.A.B. collected and analysed the pigment samples using synchrotron powder
XRD. J.S. summarized palaeoclimate data for northern Australia. L.L., M.Sa.,
M.P., G.P. and T.Mu. performed isotopic and sediment analyses. C.P. performed
skeletal analysis and assisted with in-field excavation processing. C.C., Z.J., B.M.
and R.G.R. wrote the main text with specialist contributions from other authors.
Author Information Reprints and permissions information is available at
www.nature.com/reprints. The authors declare no competing financial
interests. Readers are welcome to comment on the online version of the paper.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations. Correspondence and
requests for materials should be addressed to C.C. (c.clarkson@uq.edu.au)
and Z.J. (zenobia@uow.edu.au).
Reviewer Information Nature thanks R. Dennell, C. Marean, E. J. Rhodes and
J.-L. Schwenninger for their contribution to the peer review of this work.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
article reSearcH
METHODS
Excavation methods. Excavation took place in 1 × 1 m squares in 2–5 cm spits
within stratigraphic layers. Twenty 1 × 1 m squares were excavated adjacent to and
enclosing the original 1973 and 1989 excavations (Fig. 1c). Eight squares were dug to
2.75 m depth (B1, C1–3, D1–3, E2), while five squares were dug to 3 m depth (B5, B6,
C4–6) and two squares (B2, B3) to 3.4 m depth in 2012 and 2015. One square was dug
to 4.3 m depth in 1989 (B4), with the remaining squares discontinued at 1.3–1.5 m
depth to maintain baulks and site stability (E1, E3, E4, D4). The deeper squares were
excavated in 72 spits, with spits averaging 4–5 cm deep in the upper deposit where
dense shell midden was present, and averaging 2.5 cm deep from 2–3 m depth. All
artefacts larger than 2 cm were piece-plotted in situ with a total station and bagged
separately with a unique identifying number. Artefacts were identified during the
excavation, where possible, and coded by type (lithic, haematite, bone, human bone,
axe, axe flake, grinding stone, and so on). Sedimentary features (abbreviated as SF)
were outlined and contours plotted with a total station. The latter included human
burial cuts, hearths, pits and intrusive features such as root casts, burrows and
postholes. All sediments were passed through nested 3 and 7 mm mesh sieves
and a 100 g sediment sample was retained from each spit. Munsell colour and
pH values were obtained for each spit and all sieve residue was bagged and trans-
ported to the University of Queensland archaeology laboratories for sorting.
Sediments were extensively sampled for pollen, geoarchaeology and phytoliths,
and blocks were taken from the walls for soil micromorphology. Bulk sediment
samples were also collected from two columns (see ‘Archaeobotanical analysis
methods’). The site was backfilled with the original sediment and all human bone
was reinterred after analysis at the completion of the excavation in 2016.
Artefact analysis methods. All stone artefacts were counted and weighed
according to raw material, artefact type (core, flake, retouched flake, flaked piece)
and typology. A complete list is provided in the Supplementary Information for
select squares. The number of artefacts per spit was divided by litres excavated
to calculate artefact frequencies (density per spit). Plotted artefacts were left
unwashed. Heat-affected artefacts were identified by high lustre, crenated
fracture, crazing, irregular heat-exfoliation surfaces and pot-lid scarring.
Ground and retouched artefacts were initially inspected under low magnification
and selected items were analysed for functional traces, including usewear and
residues. Haematite pieces were classified as ground and unground, and counted
and weighed. Ground haematite was analysed in detail for the number of ground
facets, quality and colour, and selected pieces were further analysed under high
magnification for usewear and residues.
Radiocarbon dating of charcoal. Charcoal samples were submitted for analysis
to two different radiocarbon laboratories—Waikato (Wk) and ANSTO (OZ) in
New Zealand and Australia, respectively. Samples were collected during the 2012
and 2015 excavation seasons and were carefully recorded and plotted using a total
station. The samples were collected from a number of different squares (B3–E3,
C4, C5 and E4). The relevant squares are shaded in grey in Extended Data Fig. 8a
and the sample codes, contexts and chemical pretreatments are provided in
Supplementary Table 2. Samples were taken from depths of between 8 cm and
200 cm below the ground surface.
Samples submitted to Waikato were first physically cleaned of any adhering
sediment and loose material, and then crushed. This was followed by an acid–
base–acid (ABA) chemical pretreatment, in which samples were washed in hot
HCl, rinsed, treated with NaOH and rinsed again, and then treated with hot HCl,
rinsed and dried. The pretreated samples were then combusted to CO
2
by oxida-
tion at 800 °C using CuO. The CO
2
was purified in the presence of silver wire to
absorb any SOx and NOx produced. The CO2 was then reduced to graphite with
H2 at 550 °C using an iron catalyst. The pressed graphite targets were sent to the
Keck Radiocarbon Dating Laboratory at the University of California, Irvine and
the Center for Applied Isotope Studies, University of Georgia for accelerator mass
spectrometry (AMS) measurements.
Samples submitted to ANSTO were pretreated using either the ABA procedure
(for samples younger than around 20kyr) or the acid–base wet-oxidation (ABOx)
procedure (for samples older than around 20kyr). The ABOx pretreatment
included washes in acid (2 M HCl at 60 °C for 2 h), alkali (1% NaOH at 60 °C for
1 h) and acid (0.1 M K
2
Cr
2
O
7
at 60 °C for 24 h), with Milli-Q water rinses between
each step or until the solutions were clear. The pretreated samples were oven-
dried at 60 °C for two days before being combusted to CO2 using the sealed-tube
technique, after which the CO
2
was reduced to graphite using the H
2
/Fe method
41
.
A portion of graphite was used to determine the δ
13
C value, for the isotopic frac-
tionation correction, using a Vario Microcube elemental analyser and an IsoPrime
isotope-ratio mass spectrometer. AMS measurements of
14
C content were carried
out using the STAR facility at ANSTO42.
The 14C ages and related information, including age calibration performed using
the SHCal13 calibration curve43 and the OxCal version 4.2 program44, are provided
in Supplementary Table 2.
Single-grain OSL dating. OSL dating provides a means of determining the
burial ages for sediments and associated artefacts and fossils
34–36
. The time that
has elapsed since mineral grains were last exposed to sunlight can be determined
from measurements of the OSL signal—from which the equivalent dose (D
e
) is
estimated—together with determinations of the radioactivity of the sample and
the material surrounding it to a distance of around 30 cm (the environmental
dose rate). Fifty-two samples were collected for OSL dating from the upper 2.9 m
of deposit during the 2012 and 2015 excavation seasons and from four different
excavated profiles (Extended Data Fig. 8). Individual grains of quartz (180–212 μ m
diameter) were obtained from the samples and measured for their De values,
using standard procedures and tests (for example, ref. 45). To obtain an estimate
of the environmental dose rate for each sample, an internal alpha dose rate of
0.032 ± 0.010Gykyr−1 was assumed and beta dose rates were measured using
a GM-25-5 beta counter46 and the procedures described in ref. 47. Gamma dose
rates were measured directly by in situ gamma spectrometry, and cosmic-ray dose
rates were calculated using published equations
48,49
. Beta dose rates were corrected
for grain-size attenuation and the beta, gamma and cosmic-ray dose rates were
adjusted using a water content of 5 ± 2% (68.2% confidence interval) to obtain
estimates of the total environmental dose rate. The burial time of the grains in
calendar years before present is calculated as the D
e
divided by the environmental
dose rate. Age uncertainties are given at the 68.2% confidence level and were esti-
mated by combining, in quadrature, all known and estimated sources of random
and systematic error. Det ails of the preparation, measurement and analysis of single
grains, the determination of D
e
values and dose rates, and the resulting OSL age
estimates are provided in the Supplementary Information.
Independent estimates of single-grain D
e
values and dose rates were obtained
for four samples (SW13A, 11A, 7A and 5A) at the University of Adelaide by L.J.A.
In addition, the single-grain De values for four samples reported in the original
thermoluminescence dating study of Madjedbebe
13
(KTL158, 162, 164 and 165)
were re-measured and their environmental dose rates re-calculated using the
same procedures as for the other samples in this study. Details are given in the
Supplementary Information and the results are shown in Extended Data Fig. 10.
Single-grain OSL ages were put into a Bayesian statistical model on the OxCal
platform (OxCal version 4.2.4)
44,50
(Fig. 3). The samples measured from the NE
sample column (Extended Data Fig. 8e) were not included in the model, because
there is a slope from the back of the site to the front and the depth offset with the
samples collected from the SW and NW sequences is not known with sufficient
precision. The model included the two single-grain ages obtained for KTL162
and KTL164 (ref. 16)—re-calculated here using updated dose rate information
(Extended Data Fig. 10f)—and the four replicate ages for SW13A, 11A, 7A and
5A obtained independently by L.J.A.
Each OSL age was input as a C_date in calendar years before 1950 with an
associated uncertainty (the standard error of the mean). Only the random errors
(listed in parentheses in Supplementary Table 5) were included in the model,
because OSL ages do not have fully independent uncertainties51,52. As prior
information, we used the depths associated with changes in artefact technology
and stone tool raw materials, and the peaks and dips in artefact concentration
(Extended Data Fig. 2). The entire sequence was modelled as a series of seven
phases (Fig. 3), assuming that the measured ages are unordered and uniformly
distributed within a phase, and the stratigraphically lowest phase is older than those
above. A ‘boundary’ was placed at the start and end of each phase. A General t-type
Outlier Model
53
was used to assess the likelihood of each age being consistent with
the fitted model. Each age was assigned a prior outlier probability of 5%. Further
details of the Bayesian modelling, together with the modelled OSL data and model
code, are provided in the Supplementary Information.
Archaeobotanical analysis methods. Sixty litres of sediment for each excavation
unit (approximately 100% of a 4 cm spit) was collected from two columns (C3/1–27
and C2/28–57; C6/1–15 and C5/16–72) as a bulk sediment sample, allowing a
continuous sequence of archaeobotanical recovery through the deposit. In addi-
tion, all hearths and other features were collected in their entirety for flotation.
The recovery of archaeobotanical material at Madjedbebe was aided by the use of
a cascading ‘Ankara-style’ flotation tank
54,55
, which facilitated swift and compre-
hensive processing. Archaeobotanical samples were sorted and weighed. The wood
charcoal was separated from the other macrobotanical remains, both examined
separately by X.C. and S.A.F. The wood charcoal was taxonomically identified fol-
lowing the criteria of the International Association of Wood Anatomists and with
the assistance of a comprehensive wood reference collection for the region collected
by X.C. with the assistance of the George Brown Darwin Botanic Gardens56,57.
Sampling protocols were in accordance with those outlined by Asouti and Austin
58
.
The other macrobotanical remains were sorted using standard archaeobotanical
procedures. High-powered light microscopy and scanning electron microscope
imaging was used to compare the anatomical and morphological features of the
archaeological specimens to modern reference material from the region (collected
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
article
reSearcH
by S.A.F. in conjunction with the Gundjeihmi Aboriginal Corporation, Mirarr
traditional owners and the Northern Territory Herbarium). We used the criteria
proposed by Fairbairn59 and Hather60,61 to identify Pandanus sp. drupes and
vegetative parenchyma, respectively.
Artefact associations and refitting. Refitting attempts were made on all plotted
and 7-mm-sieve silcrete artefacts as well as all 3-mm-sieve silcrete artefacts from
squares C2–C4 (n = 778). A traditional approach to refitting was combined with
high resolution, objective measurements of artefact colour to determine potential
matches between similarly coloured artefacts. This method used a portable, wire-
less colour scanner (NODE+ Chroma 2.1 by Variable Inc.) that communicates
with handheld iOS and Android devices. An application program interface for
each device was produced to allow the collection of identification-tied data, and
a project-specific.net application that allowed transparent statistical analysis and
colour grouping of data was also produced62. Refits were first attempted between
all artefacts grouped by similar colour, and then across colour grades. Scans were
restricted to a section of the lithic artefact where NODE+ either sits flush with the
surface or on a slight concavity that rises to meet the aperture. This maintained a
standard measurement distance during each scan. A minimum of three suitable
scanning areas was identified on each artefact, with five scans taken from each
sample area. Five scans were taken to minimize the potential of any human error
during positioning of the device. For banded or variably coloured artefacts, each
area of colour was treated as a separate sample area. Potential matching artefacts
were analysed for similarity under 10× and 20× magnification using an Olympus
SZX16 stereoscope with NIS Elements Advanced Research version 4, following the
methodology outlined by Wilkins63.
Granulometry. Hand-grab bulk samples were taken from squares B2, C4, and E2
at 5 cm intervals from the surface, dry sieved through a 2-mm screen and macro-
scopic organic materials were removed by hand. For each sample, three trials of five
sub-samples were analysed with a Horiba LA-950 Laser Particle Size Analyser at
the University of Washington, Department of Materials Science and Engineering,
resulting in 15 measurements per sample.
Carbon isotopes. Hand-grab bulk samples were collected from square B2,
following the procedure for granulometry. A 2 g sub-sample was then ground to
a fine powder and treated with 2 M HCl for 24 h to remove inorganic carbon.
The samples were then rinsed in de-ionized water, the water separated from the
sediments using a centrifuge, and the samples dried at 60 °C for 24 h and then meas-
ured using a Costech Elemental Analyser on a Finnigan 253 Mass Spectrometer
at the IsoLab in the University of Washington, Department of Earth and Space
Sciences.
Micromorphology. Intact blocks of sediment were extracted from squares B2
and E2, and the hearth at C4/36A, by encasing the blocks in plaster bandages
to ensure their integrity. The blocks were air-dried at 40 °C for seven days, then
impregnated with Reichhold Polylite polyester resin, styrene and hardened with
methyl ethyl ketone peroxide using a ratio of 7:3:0.025. After curing for several
weeks, thick sections were cut from the blocks with a diamond saw and sent to
Spectrum Petrographics to prepare thin sections for microscopic analysis. Thin
sections were analysed under different magnifications and different lights (plane
polarized, cross-polarized and fluorescent) with stereo petrographic microscopes
at the University of Tübingen and the University of Washington. Whole-slide scans
were taken with a flat-bed document scanner.
Magnetic susceptibility. Magnetic susceptibility (χ) was measured at both low
(460 Hz) and high (4,600 Hz) frequencies for the stratigraphic units within the
sedimentary sequence in squares B2 and C3. As observed in other sandstone
rock shelters64,65, samples are weakly magnetic in the culturally sterile layers.
The lower susceptibility values measured in the deepest deposits were often close
to the sensitivity limit of the Bartington Instrument MS2B sensor, resulting in a
higher percentage loss of the low-frequency measurements (χfd%), with averages
of around 16%; these data were discarded.
Pigment characterization. X-ray fluorescence. Non-invasive elemental charac-
terizations were undertaken to investigate the inorganic chemical composition
of pigments. Data were collected using a Bruker S1 Titan 800 portable X-ray
fluorescence (pXRF) instrument, equipped with a silicon drift detector, Rh target
X-ray tube (maximum voltage 50 kV, default to 150 °C with ultralene window)
and five-position motorized filter changer. Two beam phases were run sequen-
tially, each collecting for 90 analytic seconds. Phase 1 parameters: 45 kV, 10.45 μ A
with a filter (Ti 25 μ m, Al 300 μ m) in the beam path. Phase 2 parameters: 15 kV,
31.55 μ A without a filter. Spectra were collected on suitable artefact surfaces, where
attenuation could be minimized or avoided
66
. Relative abundance concentrations
for 27 elements reported derive from the manufacturer’s fundamental parameters
calculation.
Synchrotron powder diffraction. Powder pigments were collected using a micro-
drill, two from ART9 with an additional 75 samples taken from ochre nodules by
milling the outside of the specimens and then drilling to a depth of around 5 mm
to sample their internal structure. Powders were homogenized (manually ground
with mortar and pestle), placed into 0.3-mm-diameter borosilicate capillaries
and mounted on the beamline. Diffraction data were collected at the Australian
Synchrotron at a wavelength of 0.77412 Å, calibrated using a NIST SRM 660b,
from 5–85° 2Theta, with a Mythen microstrip detector with an inherent step size
of 0.002°, using two detector positions and a collection time of 5 min per position.
Samples were rotated at around 1 Hz during data collection to ensure good pow-
der averaging. Phase identifications of selected samples were undertaken using
Panalytical Highscore with the ICDD PDF4 database.
Usewear and residues. All potential grinding stones and fragments (n = 91) from
the 2012 excavations were examined for wear and residue traces. Stones were vis-
ually scanned under low and high magnification on the ground and unground
surfaces, using stereo and metallographic microscopes with low-angled oblique
light and vertical incident light sources, respectively. The unground surfaces were
documented to evaluate residues and traces that might be linked to handling or
anvil positioning during use, and to identify traces that mimic usewear, such as
micro-fractures of quartz grains on non-used surfaces. The latter are probably
caused by friction between sediment and artefacts, either during use or after dis-
card and burial within the sediments. All complete, near-complete and broken
edge-ground hatchets (n = 10), and a selection of flakes from edge-ground axe mar-
gins, were also examined under the same stereo and metallographic microscopes.
Residue distributions were documented across each of the tool surfaces and
characterized using a range of techniques, following removal with a solvent mixture
of water, ethanol and acetonitrile (up to 50 μ l extracted with an adjustable pipette).
Residue extractions were characterized using: (1) high-magnification transmitted
light microscopy and various biochemical stains to test for and distinguish plant
and animal tissues; (2) absorbance spectroscopy and biochemical tests for the
detection of protein, carbohydrates, fatty acids, starch and ferrous iron (see ref. 67
and Supplementary Information for specific methods for each test); and (3) gas
chromatography mass spectrometry. A selection of grinding stones (n = 12) had
additional residues removed by ultrasonication. Tools were completely or partially
submerged in distilled water and ultrasonicated for 2 min. Density separation of
the extracted residue mixtures isolated starch grains and other plant microfossils
(phytoliths, raphides, pollen), when present.
Code availability. The computer code used to generate the Bayesian age model
for the site is provided in full in the Supplementary Information, together with
information about the program and version used. The R code used to analyse and
visualize the geoarchaeological and stone artefact data is archived online at https://
doi.org/10.6084/m9.figshare.4652536.
Data availability. All elements necessary to allow interpretation and replication
of results, including full datasets and detailed experimental procedures are
provided in the Supplementary Information. All geoarchaeological and stone
artefact data files are archived online at https://doi.org/10.6084/m9.figshare.
4652536. Archaeological material generated in this study will be kept in the
Archaeology Laboratories of the University of Queensland until 2018. It will then
be deposited in a Gundjeihmi Aboriginal Corporation keeping place. The material
will be publicly accessible upon request with permission from Gundjeihmi
Aboriginal Corporation and the corresponding author. Archaeological materials
from the 1973 and 1989 excavations are stored in the Museum and Art Gallery
of the Northern Territory and can be publicly accessed with permission from the
museum director.
41. Hua, Q. et al. Progress in radiocarbon target preparation at the Antares AMS
Centre. Radiocarbon 43, 275–282 (2001).
42. Fink, D. et al. The ANTARES AMS facility at ANSTO. Nucl. Instrum. Methods Phys.
Res. B 223–224, 109–115 (2004).
43. Hogg, A. G. et al. SHCal13 Southern Hemisphere calibration, 0–50,000 years
cal BP. Radiocarbon 55, 1889–1903 (2013).
44. Bronk Ramsey, C. & Lee, S. Recent and planned developments of the program
OxCal. Radiocarbon 55, 720–730 (2013).
45. Wood, R. et al. Towards an accurate and precise chronology for the colonization
of Australia: the example of Riwi, Kimberley, Western Australia. PLoS ONE 11,
e0160123 (2016).
46. Bøtter-Jensen, L. & Mejdahl, V. Assessment of beta dose-rate using a GM
multicounter system. Int. J. Rad. Appl. Instrum. D 14, 187–191 (1988).
47. Jacobs, Z. & Roberts, R. G. An improved single grain OSL chronology for the
sedimentary deposits from Diepkloof Rockshelter, Western Cape, South Africa.
J. Archaeol. Sci. 63, 175–192 (2015).
48. Prescott, J. R. & Hutton, J. T. Cosmic ray contributions to dose rates for
luminescence and ESR dating: large depths and long-term time variations.
Radiat. Meas. 23, 497–500 (1994).
49. Smith, M. A., Prescott, J. R. & Head, M. J. Comparison of 14C and luminescence
chronologies at Puritjarra rock shelter, central Australia. Quat. Sci. Rev. 16,
299–320 (1997).
50. Bronk Ramsey, C. Bayesian analysis of radiocarbon dates. Radiocarbon 51,
337–360 (2009).
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
article reSearcH
51. Rhodes, E. J. et al. Bayesian methods applied to the interpretation of multiple
OSL dates: high precision sediment ages from Old Scatness Broch excavations,
Shetland Isles. Quat. Sci. Rev. 22, 1231–1244 (2003).
52. Galbraith, R. F. & Roberts, R. G. Statistical aspects of equivalent dose and error
calculation and display in OSL dating: an overview and some
recommendations. Quat. Geochronol. 11, 1–27 (2012).
53. Bronk Ramsey, C. Dealing with outliers and osets in radiocarbon dating.
Radiocarbon 51, 1023–1045 (2009).
54. French, D. H. An experiment in water-sieving. Anatol. Stud. 21, 59–64
(1971).
55. Nesbitt, M. Plants and people in Ancient Anatolia. Biblic. Archaeol. 58, 68–81
(1995).
56. Wheeler, E. A., Baas, P. & Gasson, P. E. IAWA list of microscopic features for
hardwood identication. IAWA Bull. 10, 219–332 (1989).
57. Richter, H. G., Grosser, D., Heinz, I. & Gasson, P. E. IAWA list of microscopic
features for softwood identication. IAWA J. 25, 1–70 (2004).
58. Asouti, E. & Austin, P. Reconstructing woodland vegetation and its exploitation
by past societies, based on the analysis and interpretation of archaeological
wood charcoal macro-remains. Environ. Archaeol. 10, 1–18 (2005).
59. Summerhayes, G. R. et al. Human adaptation and plant use in highland
New Guinea 49,000 to 44,000 years ago. Science 330, 78–81 (2010).
60. Hather, J. G. in Tropical Archaeobotany: Applications and New Developments
(ed. Hather, J. G.) 51–64 (Routledge, 1994).
61. Hather, J. G. Archaeological Parenchyma (Archetype Publications, 2000).
62. McNeil, J.-L., Marginson, A., Mackay, A. & Clarkson, C. Colour Signature
Analysis: Using objective colour quantication techniques towards retting
lithic assemblages. 80th Annual Meeting of the Society for American
Archaeology (2015).
63. Wilkins, J. R. Prepared Core Technology at Kudu Koppie, South Africa and the
Modern Human Behaviour Debate. MA thesis, Univ. Calgary (2008).
64. Lowe, K. M. et al. Using soil magnetic properties to determine the onset of
Pleistocene human settlement at Gledswood Shelter 1, northern Australia.
Geoarchaeology 31, 211–228 (2016).
65. Lowe, K. M., Mentzer, S. M., Wallis, L. A. & Shulmeister, J. A multi-proxy study of
anthropogenic sedimentation and human occupation of Gledswood Shelter 1:
exploring an interior sandstone rockshelter in Northern Australia. Archaeol.
Anthropol. Sci. http://dx.doi.org/10.1007/s12520-016-0354-8 (2016).
66. Huntley, J. Taphonomy or paint recipe? In situ portable X-ray uorescence
analysis of two anthropomorphic motifs from the Woronora plateau, New
South Wales. Aust. Archaeol. 75, 78–94 (2012).
67. Fullagar, R. et al. Evidence for Pleistocene seed grinding at Lake Mungo,
south-eastern Australia. Archaeol. Oceania 50, 3–19 (2015).
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
article
reSearcH
Extended Data Figure 1 | Madjedbebe rock shelter. a, Section drawing
of the southwest profile wall, showing major stratigraphic divisions and
sediment descriptions, and the location of the 1973, 1989, 2012 and 2015
excavation trenches. Light grey dots show plotted artefacts. b, Photograph
of the site during the 2015 excavation. c, Detail of the site ground surface
during ground penetrating radar survey, before the 2012 excavation.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
article reSearcH
Extended Data Figure 2 | Plot of artefact densities and assemblage
composition as a function of depth below ground surface. a, Plot of
density of artefacts found during the 2012 and 2015 excavation seasons
in squares from the C and B rows. Artefacts are shown by type (axe flake,
ochred slab, axe or axe fragment, grinding stone, ground ochre, and flake
or core) superimposed on the southwest profile wall (Extended Data Fig. 1).
Phases represent the three dense artefact bands (see text and Supplementary
Information). b, Plot of artefact density and raw material type with depth,
based on plotted artefacts and residue found in the 7-mm sieves for square
B6. c, Plot of technological changes with depth, based on plotted artefacts
and residue found in the 7-mm sieves for square B6.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
article
reSearcH
Extended Data Figure 3 | Grinding stones, residues and usewear
of specimens collected from phase 2 at Madjedbebe. a–f, Specimen
UPGS36 (from 2012 spit 44) and residues from processing of red pigment.
a, Ground surface. Scale bar, 2 cm. b, Plan view. c, Ground surface at low
magnification (location 1 in a) showing levelled grains. d–f, Red pigment
residues at high magnification. d, Location 2 in a. Scale bar, 0.5 mm.
e, f, Location 2 in a. Scale bars, 0.02 mm. g–k, Specimen GS39 (from 2012
spit 37) and usewear, used for processing of seeds. g, Ground surface. Scale
bar, 4 cm. h, Plan view. i, Ground surface at low magnification (location 1
in g) showing levelled and rounded grains. j, Bright use-polish with
striations (arrows, location 2 in g). Scale bar, 0.1 mm. k, Bright, reticulated
use-polish (location 3 in g). Scale bar, 0.05 mm. l, Specimen GS73 (from
2015 spit 52): bright, undulating use-polish, with red pigment residues in
the lowest regions of the grains (circle, location 1 in s). Scale bar, 0.05 mm.
m–r, Specimen GS79 (from 2015 spit 54) used for the manufacture and
sharpening of stone hatchets. m, Plan view. Scale bar, 5 cm. n, Ground
surface. o, Side view. p, Angled view, upper surface is ground, note the
flake margins. q, Location 2 in p showing flake scars. r, Ground surface
at low magnification (location 1 in n) showing levelled grains and deep
striations (arrows). s–v, Fragment of GS73 with deep partial grooves:
s, Ground surface. Scale bar, 5 cm. t, Side view. u, Plan view. v, Ground
surface at low magnification, note the deep striations and red surface
staining (location 2 in u).
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
article reSearcH
Extended Data Figure 4 | Dolerite edge-ground hatchet heads showing
plan and end views. Main scale bars are 5 cm. Vertical double-ended
arrows indicate the haft zones. a, EGH7 from unit C1/35 (base of phase 3)
with shouldered or stemmed design for a haft. Two upper insets show
(left; scale bar, 2 mm) striations from grinding and (right; scale bar,
0.2 mm) polish from use. The lower insets show (left; scale bar, 0.2 mm)
wear from haft movement and (right; scale bar, 0.01 mm) detail of the
polish (smooth white zones) and possible resin (red smears with black
spots). b, EGH1 from unit C1/33 (phase 4) with large flake scarring and
cracks within the haft zone. c, EGH8 from unit C1/38 (base of phase 3)
with a slight waist design for a haft. d, EGH6 from C1/33 (phase 3) with
grooved design for a haft and red stain from mixing pigment (ellipse). The
upper inset (scale bar, 2 mm) shows traces of use (vertical arrows) and
grinding (horizontal arrows). The lower inset (scale bar, 0.2 mm), from
inside the groove, shows polish from haft movement.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
article
reSearcH
Extended Data Figure 5 | Hearth SF56 with grindstones and carbonized
Pandanus drupe from a hearth in spit C2/41. a, Photograph of hearth
pit SF56 in C4/35 (phase 3) showing in situ grinding stones in a hearth
with elevated magnetic susceptibility readings, and a probable cache of
ground ochre, grindstones and hatchet heads against the back wall.
b, d, Scanning electron microscope images of modern reference specimen
2639, Pandanus spiralis drupe (13× and 90× magnifications, respectively).
c, e, Photographs of archaeological specimen C2/41(1), Pandanus sp.
drupe. Note the seed locule, vascular bundles and flaring ground tissue
apparent on both modern reference and archaeological specimens.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
article reSearcH
Extended Data Figure 6 | Summary of Madjedbebe silcrete artefact
refitting analysis. a, Selection of refitting and conjoining artefacts; scale
bar intervals, 10 mm. b, Histogram showing the distribution of vertical
distances between refitting artefact fragments. The median vertical
refit distance is 0.10 m, with a median absolute deviation of 0.13 m.
c, Histogram showing the distribution of straight-line distances between
refitted artefact fragments. The median straight-line refit distance is
0.44 m, with a median absolute deviation of 0.47 m. d, Plan view showing
the refitted artefacts at the locations where they were found at the time
of excavation. Blue lines connect refitted pieces. Annotations on the
axes show the excavation grid coordinates. e, Polar plot of horizontal
orientations of the vector between pairs of refitted pieces. The Rayleigh
test result indicates a significantly non-random distribution. For most
refits, both artefacts in the refit pair were recovered from the same
horizontal plane. f, Section view showing the refitted artefacts at the
locations where they were found at the time of excavation. Blue lines
connect refitted pieces. g, Plot of artefact mass by depth in square B6: each
point represents one artefact, the blue line is a robust locally weighted
regression, and the grey band is the 95% confidence region for the
LOWESS regression line.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
article
reSearcH
Extended Data Figure 7 | See next page for caption.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
article reSearcH
Extended Data Figure 7 | Summary of Madjedbebe geoarchaeological
analysis. a, Particle size distributions of bulk samples extracted from
the southwest wall of square D3 (left) and constrained cluster analysis
dendrogram (right). Blue horizontal lines indicate the artefact discard
phases, calibrated for squares C3 and D3. b, Distributions of key
geoarchaeological variables measured on bulk samples extracted from
the southwest wall of square D3. Magnetic susceptibility units are
10−7 m3kg−1; VPDB is Vienna Pee Dee Belemnite, an international
reference standard for δ
13C analysis. c, Scanning electron microscope
images of sand grains from 1.35 m (top) and 3.20 m (bottom) depth below
surface (bs). d, Photograph of the northeast section of the 2012 excavation
area. Labels in white circles indicate locations of micromorphology
samples. e, Micromorphology sample NE1 from the midden deposit
showing shell fragment (red arrow), charcoal (green arrow) and root
fragment (blue arrow). f, Micromorphology sample NE2 from the lower
midden deposits showing linked-capped grains (red arrow), silt (blue
arrow) and voids (green arrow). g, Micromorphology sample NE3 from
below the midden showing weathered charcoal fragment with clay infill
(red arrow). h, Micromorphology sample NE4 showing an extensively
weathered charcoal fragment. i, Micromorphology sample NE5 showing
grain with silty coating (red arrow), grain with clay coating (blue arrow)
and grain with no coating (green arrow). j, Micromorphology sample from
the C2/36 hearth feature showing a well-preserved charcoal fragment.
k, Micromorphology sample from the southwest section of square D3
(2.18–2.25 m depth below surface) showing linked-capped grains
(red arrows), similar to sample NE2. l, Micromorphology sample from
the southwest section of square D3 (2.22–2.29 m depth below surface)
showing packing voids (green arrow) and a polymineral grain with linked-
capping joining it with smaller grains (red arrows).
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
article
reSearcH
Extended Data Figure 8 | Site plan, OSL sample locations and ages.
a, Two-dimensional site plan of excavated squares, showing the locations
of the OSL sample series. Grey-shaded squares represent squares
from which charcoal samples were collected for 14C dating. b, Three-
dimensional site plan, showing both horizontal and vertical positions
of the OSL sample series. Samples shown in the same colour were taken
from section walls with the same orientation. c–f, Photographs of the
sedimentary deposit for each of the walls from which OSL samples were
collected, together with the OSL ages (uncertainties at 68.2% confidence
level) and the lowest dense artefact band (phase 2) demarcated by the
stippled lines. c, Southwest wall of square B5. d, Southwest wall of
square B6. e, Northeast wall of square E2. f, Northwest wall of square C4.
g, Comparison of 14C and OSL ages (uncertainties at 95.4% confidence
level) obtained in this study from the upper 2 m of deposit.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
article reSearcH
Extended Data Figure 9 | OSL data for individual sand-sized grains
of quartz. a–k, Radial plots of single-grain De values for each sample
within the lowest dense artefact band (phase 2). a, SW4C; b, SW3C;
c, NW14; d, SW13A; e, SW2C; f, NW13; g, SW11A; h, NW12; i, NW11;
j, SW10A; k, NW9B. l, Radial plot of De values for single grains of
sample NE1, collected from the shell midden at the top of the sequence.
The grey bands in each plot are centred on the weighted mean De
determined for each dose population using the central age model, after
the rejection of outliers (shown as open triangles). m, OSL decay curves
for a representative sample of grains from SW13A that span the range of
observed luminescence sensitivities (that is, their relative brightness). The
inset plot shows the same curves on a normalized y axis. n, Corresponding
dose response curves for the grains shown in m.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
article
reSearcH
Extended Data Figure 10 | See next page for caption.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
article reSearcH
Extended Data Figure 10 | Verification of previous luminescence ages
and inter-laboratory comparison. a, Schematic diagram of square B4
(modified after ref. 13) showing the relative positions of four samples
for which ages have been reported previously13,15,16 and that were
re-measured and evaluated in this study. b–e, Radial plots of single-grain
De values measured in this study for these four samples. b, KTL165;
c, KTL164; d, KTL158; e, KTL162. The grey bands in each plot are
centred on the weighted mean De determined for each dose population
using the central age model, after the rejection of outliers (shown as open
triangles). f, Previously published De values, total dose rates and ages,
together with the revised dose rates and ages (values in parentheses; see
Supplementary Information for explanation) and the new single-grain
OSL De values (based on the data shown in b–e) and ages obtained in
this study. g–j, Radial plots of single-grain De values for the four samples
measured independently in two laboratories (University of Wollongong,
UOW; University of Adelaide, UA). g, Sample 1 (SW13A); h, Sample 2
(SW11A); i, Sample 3 (SW7A); j, Sample 4 (SW5A). Filled circles and
open triangles are De values obtained at UA and UOW, respectively. The
grey bands in each plot are centred on the weighted mean De determined
using the central age model for each dose population measured at UA.
k, Comparison of weighted mean De and overdispersion (OD) values
for the same samples measured at UA and UOW (‘A’) using a preheat
combination of 260 °C for 10 s (PH1) and 220 °C for 0 s (PH2), and at UOW
(‘B’) using a preheat combination of 220 °C for 10 s (PH1) and 160 °C for
5 s (PH2). l, High-resolution gamma-ray spectrometry results obtained
at UA and the beta and gamma dose rates and OSL ages calculated from
these data, compared to the beta and gamma dose rates and OSL ages
obtained independently at UOW (using preheat combination ‘A’ for De
determination).
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.