Journal of Archaeological Science 32 (2005) 167–181
http://www.elsevier.com/locate/jas
Archaeobotany of fruit seed processing in a monsoon savanna
environment: evidence from the Keep River region, Northern
Territory, Australia
Jennifer Atchisona, Lesley Heada,*, Richard Fullagarb
a
GeoQuEST Research Centre, School of Earth and Environmental Sciences, University of Wollongong, Wollongong 2522, Australia
b
Department of Prehistory and Historic Archaeology, University of Sydney, 2000, Australia
Received 15 April 2003; received in revised form 11 March 2004
Abstract
We analyse archaeobotanical remains from three excavated rockshelter sites, Jinmium, Granilpi and Punipunil, in the Keep
River region, northwestern Australia. The record is dominated by burnt fragmented seed remains from the fruit trees Persoonia
falcata and Buchanania obovata, consistent with ethnographic records of whole fruits being pounded into pastes and cakes at the
beginning of the summer wet season. Surface seed samples of non-cultural origin are mostly whole and unburnt, and contain higher
proportions of grass seeds. Sustained processing of fruit seeds is first visible in the archaeological record about 3500 years ago.
Spatial and temporal variation in its intensity is evident since that time until it declines following European colonisation. The decline
does not represent total site abandonment, but a reorientation of activities following the ecological and social changes that came
with pastoralism. The former included the local decline of P. falcata with more intense fire regimes.
Ó 2004 Elsevier Ltd. All rights reserved.
Keywords: Persoonia falcata; Buchanania obovata; Ethnobotany; Holocene
1. Introduction
Hunter–gatherer interactions with plants have long
been an important ingredient of debates surrounding
agricultural origins [12]. As Hastorf [13] has argued,
there are now many archaeological examples of plant
cultivation and use either preceding, or independent of,
the morphological changes in plants usually involved
in strict definitions of agriculture. Many plants were
actively tended, moved around and actively used for
thousands of years without visible morphological
* Corresponding author. Tel.: C61 2 4221 3124; fax: C61 2 4221
4250.
E-mail address: [email protected] (L. Head).
0305-4403/$ - see front matter Ó 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jas.2004.03.022
change ([13]: p. 773). This significant problem of
archaeological visibility continues to constrain understanding of the range of ways people interacted with
their environments during the Holocene.
The focus on morphological change has involved
a focus on grains and starchy staples (e.g. [19]) and their
role in the reproduction of sedentism, surplus and
hierarchy. However, Hastorf ([13]: p. 773, after Farrington and Urry [9]) argues that the earliest propagated
plants included rather the ‘medicinal, industrial, spicy,
hallucinatory or merely exotic’. In this paper we use the
example of fruits and fruit processing. Although their
archaeological preservation is somewhat fortuitous (cf.
[4]), these remains illustrate a broader social context of
plant use in which preservation, storage, seasonality and
gender are all important variables.
168
J. Atchison et al. / Journal of Archaeological Science 32 (2005) 167–181
In her overview Hastorf draws most examples from
South America, but the issues are salient in a variety
of environmental contexts. Despite an extensive ethnobotanical literature relevant to Aboriginal use and
processing of plant foods in northern Australia, the
archaeobotanical visibility of these activities remains
limited. Aboriginal people maintain detailed knowledge
of – and in many cases practise – the management and
utilisation of carbohydrate staples, seasonally abundant
fruits, grass seeds and potentially toxic plant foods such
as cycads (e.g. [8,16,18,21,28,31]).
The seasonal tropics offer particular challenges to
archaeological preservation. Microscopic material preservation sites, notably pollen, are restricted to the
north east of the continent, and most sites with
archaeobotanical remains are restricted to the Holocene [5,10,17,20,25,30]. Consequently any identifiable
material recovered in the region is of significance. On
a continental scale the picture of Australian Aboriginal plant interactions provided by archaeobotany is
at best fragmentary, containing poor quality preserved
material in mostly localised recorded sites. A notable exception is the Carpenter’s Gap site in the
western Kimberley, dated to about 40 000 BP [23,24].
McConnell and O’Connor examined extensive macrobotanical remains, inferring continued Aboriginal
occupation and utilisation of a wide variety of species
in the southwestern Kimberley during the last glacial
maximum.
In this paper we analyse new material from archaeobotanical remains at three sites across a region of the
eastern Kimberleys: Jinmium, Granilpi, and Punipunil
excavated in the Keep River, northwest Northern Territory (Fig. 1). This study is part of a broader project
examining long term relationships between Aboriginal
people and their environment in the region. Our specific
aims here are to:
distinguish between the cultural and non-cultural
components of the archaeobotanical record via
morphology and species assemblage,
examine the chronology and taphonomy of the
cultural plant remains,
by comparing archaeobotanical and ethnobotanical
evidence, infer patterns of Aboriginal plant utilisation and environmental interaction both prior to and
immediately after European contact.
We will argue that plant visibility in this archaeological record is a direct result of people processing
the fruit, adding complexity to questions of resource
abundance and availability. We will also demonstrate
that the plant record contributes significant information
to broader understandings of hunter–gatherer response
to change by adding detail to questions about seasonality, resource manipulation and burning.
Fig. 1. Keep River study area, northwestern Northern Territory,
Australia.
2. Study area
The areal focus of the study is the lower Keep River
catchment in the northwest corner of the Northern
Territory (Fig. 1). The region has a warm, dry monsoonal climate with a distinct wet season between
December and April, and annual rainfall of 750–
900 mm. Most important land systems in the area are
estuarine deltaic plains, open woodlands over tall grass
sandy plains, and rugged sandstone and conglomerate
hills with a series of more isolated outliers. Jinmium,
Granilpi and Punipunil are all examples of the latter.
Aboriginal groups with attachments to this area include
Gajerrong, Miriuwong, Jamanjung and Murinpatha
people.
The region is dominated by a large sandy plain,
predominantly vegetated by savanna woodland. The
upper stratum of this savanna is mostly Eucalyptus with
Eucalyptus tetradonta and Eucalyptus miniata being
dominant. Other common genera include Acacia,
Terminalia and Gardenia. The understorey is mostly
grass with Sehima, Chrysopogan, Themeda, Heterpogan,
Sorghum and Plectrachne being common [29]. Within
the savanna woodlands a number of trees that produce
edible fruits are also present. We focus here on
Buchanania obovata (Murinpatha name: kilen, common
names: bush mango, wild mango, green plum) and
Persoonia falcata (Murinpatha name: kathan, common
name: wild pear). B. obovata trees are today scattered
widely throughout the savanna, but also occur in
J. Atchison et al. / Journal of Archaeological Science 32 (2005) 167–181
clusters that offer particularly abundant sources of fruit.
P. falcata now has a much more restricted distribution in
the study area, for reasons discussed later in the paper.
3. Methods
Of the dozen excavations undertaken for the project,
those with significant quantities of archaeobotanical
remains including consistent preservation at depth were
chosen for detailed analysis. Some excavated pits
contained very small volumes of poorly preserved
material that were not useful for identification and
analysis. Recovered material from the seven suitable
1 ! 1 m pits (four at Jinmium, one at Granilpi and two
at Punipunil) are discussed here. Excavations followed
natural stratigraphic changes with a maximum spit
depth of 15 cm. All excavated material was dry sieved in
the field through 2–4 mm and O4 mm sieves, prior to
being transported back to the laboratory. Bulk samples
of the !2 mm sediment were also collected.
An assessment of the modern seed assemblage close
to each archaeological site can provide a useful picture
of contemporary seed distribution. This analogue enables an examination of the present day vegetation,
contemporary dispersal processes, and resulting modern
assemblages, as well as some insight into what differences there might be between cultural and non-cultural
assemblages. Seed distribution is influenced by a range
of dispersal factors including, distance from parent
plant, wind, rolling, and animal action such as birds,
bats etc. Nine modern samples were collected at
Jinmium and seven at Marralam over a 2 ha plot
containing the excavated pits. This included sampling
under the entire range of parent tree species present at
each site. An area of soil and surface litter approximately 20 cm ! 30 cm (the size of a spade) and 10 cm
deep was collected in each case. For consistency with the
excavation samples the surface samples were also sieved
back in the laboratory using the same sieve sizes (4 mm
and 2 mm).
All samples were extremely dry and fragile, with high
proportions of burnt and carbonised material. Experiments with wet sieving [3] caused considerable damage
to the remains so only hand sorting was undertaken.
Material from both surface and archaeological samples
was sorted in two size fractions, O4 mm and 2–4 mm. A
pilot study of seed recovery from the !2 mm fraction of
the surface samples indicated that the time and expense
of this procedure was not warranted by the additional
information obtained. In 15 of the 16 samples more
species were recovered from the O2 mm than !2 mm
fractions, so the smaller fraction of the archaeological
sediments was not analysed.
The lack of archaeobotanical keys in Australia [6]
means that reference collections need to be developed
169
for individual projects. A collection of representative
botanical material prepared by Allen [2] and additional
material obtained by JA were sent to the Parks and
Wildlife Commission of the Northern Territory herbarium for identification. Further specimens (predominantly grasses) were sent to the N.S.W. Department of
Agriculture Seeds Laboratory, the Australian National
Herbarium and the Queensland herbarium of the
Brisbane Botanic Gardens. Identified and unidentified
seeds were counted, photographed and examined for
morphological evidence of gnawing or other use–wear
features and compared with recorded ethnographic
information for evidence of processing. Through these
methods a significant proportion of the collection was
identifiable, however quantities of unidentified specimens are also recorded in the results. The size of the
reference collection is small and constrained as wet
season access to the sites was not possible. This limited
the flowering and fruiting bodies that could be collected
from each species, however the NT government collections are extensive and were utilised with the
assistance of trained botanists.
B. obovata seed fragments can be distinguished from
P. falcata seed fragments on the basis of surface patterning and shape. B. obovata seeds showed some surface patterning on the outer surface and in most cases
had fractured along the natural fracture plane
(Fig. 2a,b). P. falcata seeds and seed fragments had
a distinct lack of surface patterning and the seeds had no
Fig. 2. Seed remains recovered from the Keep River archaeological
excavations. a, Whole Persoonia falcata seed; b, whole Buchanania
obovata seed; c, B. obovata fragments; d, P. falcata fragments. Note
each scale division equals 1 mm.
170
J. Atchison et al. / Journal of Archaeological Science 32 (2005) 167–181
natural fracture plane. Seed fragments of P. falcata were
generally more angular than those of B. obovata
(Fig. 2c,d).
AMS 14C dating was used here to examine directly
the age of seed fragments (P. falcata) recovered from the
three archaeological deposits, for comparison with other
14C, TL and OSL dates. As well preserved P. falcata
seeds were found consistently throughout the profiles of
each of the deposits, dating was focused on this species.
4. Results
4.1. Surface seed assemblages
The common feature of all the surface assemblages at
Jinmium and Marralam is the widespread occurrence of
grass seeds, notably Heteropogan contortus and Panicum
sp. (Table 1). Fruit seeds are locally abundant close to
parent trees, for example under Erythroxylum ellipticum,
B. obovata and Terminalia latipes trees. Most recovered
material was unburnt. Samples collected from Marralam had the highest proportions of burnt to non-burnt
seeds, no doubt because the site had been burnt fairly
recently prior to sampling.
points and flaked stone, as well as other materials of
cultural origin including ochre, charcoal and various
types of bone.
In the upper Units (6 and 7) of pits C1/I (Fig. 3), C1/
II and C1/III (Fig. 4), the O4 mm fraction contains a
diversity of taxa including grasses (e.g. H. contortus),
and fruit seeds such as B. obovata, P. falcata, T. latipes,
and Vitex glabrata. In all cases the fruit seeds are
unburnt. Some whole, i.e. intact, but burnt B. obovata
and P. falcata seeds were recovered from Unit 3 of C1/I.
Seed dates are shown in these diagrams and summarised
in Table 2.
The species assemblage of the upper Units of the 2–
4 mm fractions (Figs. 5 and 6) is fairly similar with the
addition of smaller seed species such as Glycine sp.,
Setaria apiculata and Yakirra majuscula. The lower
Units of the 2–4 mm fractions are however markedly
different to the larger fractions. Most notable are the
high densities of P. falcata and some B. obovata seed
fragments, particularly in CI/III. In pit C1/I the density
of both of these species peak in Unit 3, while in pits C1/
II and C1/III these seed fragments peak in Unit 4. All of
the seed fragments from these two species below unit 5
were burnt.
4.3. Granilpi
4.2. Jinmium
The prominent stacks and boulders called ‘Jinmium’
are located at the northwestern edge of several hectares
of outcropping sandstone with rock art and archaeological deposits in at least eight rockshelters. The C1/IV
pit is located 15 m west of the main rock shelter. The
seed assemblage in this pit is wholly confined to the
upper unit, i.e. the top 40 cm and was more similar to
the surface grab samples than any of the other excavated
pits. A range of grasses was found, mostly within the
top 10 cm. Panicum sp., Mimosaceae, Lotus sp. and
P. falcata seeds were found below 30 cm, however, these
were all recovered in very low densities. A small peak
in burnt seeds is recorded in C1/IV, comprised wholly of
P. falcata seeds. While this pit contains abundant other
archaeological material, e.g. stone artefacts, it has less
culturally deposited plant material. This could be either
because processing was concentrated closer to the
shelter, or because the more open context of C1/IV on
the sand plain has mitigated against preservation of
plant material.
The maximum depth of each of the four Jinmium
excavation pits (CI/I–IV) was approximately 200 cm with
cultural material found to a depth of 150 cm. Each pit
was predominantly comprised of loose partially sorted
sand with rubble increasing in density towards the base.
Seven stratigraphic Units have been defined and are
summarised in detail in Fullagar et al. [11]. Archaeological materials included pounding stones, cores, stone
The Granilpi excavation site is within a boulder
cluster on the western side of an extensive sandstone
outcrop displaying abundant rock art [32]. The rockshelter is composed of two large leaning boulders
forming a shaded tunnel. The maximum depth reached
in the G1 pit was approximately 150 cm with cultural
material found down to this depth. One major
stratigraphic break was observed in the G1 profile
between dark organic sediment with charcoal, at 60–
70 cm, which overlays an orange sand. Archaeological
material includes high densities of flaked quartz,
quartzite and fine grained stone as well as a number of
stone points, some bone, shell and ochre.
The top few spits (Unit 3 and top half of Unit 2)
(Figs. 7, 8) of both size fractions contain a diversity of
fruit tree seeds and seed fragments, including B. obovata,
Owenia vernicosa, P. falcata, Pavetta brownii, T. latipes
and V. glabrata. H. contortus and unidentified seeds
make up the remainder of the upper assemblage. Similar
to Jinmium, the two species B. obovata and P. falcata
form the dominant component of the lower half of
Unit 2. B. obovata peaks in spit 8 (32.6–37.5 cm) and
P. falcata in spit 11 (48.75–54.1 cm). Almost all of the
seeds recovered from below 30 cm depth were burnt.
4.4. Punipunil
Punipunil is a shallow rockshelter up to 10 m from
dripline to rear rockwall lying along the base of a steep
Table 1
Numbers of whole, fragmented, burnt and unburnt seeds from the surface sampling at Jinmium and Marralam
Taxa
J1
J2
Acacia sp.
w.6, ub
w.4, ub
Digitaria sp.
Erythroxylum
ellipticum
Eucalyptus
tetradonta
Euphorbia sp.
Glycine
clandestina
Heteropogan
contortus
Lotus sp.
Opelia
amentaceae
Owenia
vernicosa
Panicum sp.
J4
J5
J6
J7
J8
w.1, ub
J9
M10
w.1, ub
M11
M12
f.3, ub
M13
M14
w.2, ub
w.35, ub;
f.25, ub
M15
M16
w.1, ub
f.8, ub
w.2, ub
w.1, ub
f.1, ub
f. 1, b
w.5, ub
w.2, ub;
f.3, ub
w.26, ub; w.6, ub;
f.8, ub
f.24, ub
f.1, b
f.1, ub;
w.1, ub
w.3, ub;
f.5, ub
w.13, ub;
f.2, ub
f.1, ub
w.12, ub; f.4, ub
f.41, ub
f.1, ub
w.1, ub
w.12, ub;
f.54, ub
f.88, ub
w.108, ub;
f.47, ub
w.1, ub
w.1, ub; w.1, ub;
f.1, ub
f.1, b
w.2, ub w.2, b
w.58, ub; w.65, ub; w.1, ub
f. 81, ub f.56, ub
w.1, ub
w.1, ub
w.8, ub;
f.1, ub
w.5, ub;
f.1, ub
w.1, ub
w.1, ub;
f.1, ub
w.140, ub w.55, ub; w.10, ub
w.2, b
w.14, ub w.27, ub
w.3, ub;
f.3, ub
w.1, ub
f.2, ub
f.4, ub
w.1, ub
w.9, ub;
w.1, ub; w.10, ub;
f.109, ub; f.33, ub f.129, ub
f.1, b
w.5, ub;
f.110, ub
w.6, ub;
f.215, ub
w.2, ub; f.69, ub
f.23, ub
f.28, ub
w.1, ub; f.16, b;
f.32, ub f.12, ub
f.41, b;
w.69, ub
f.57, b
w.30, ub; f.26, ub;
f.57, b
f.17, b
f.40, b
f.37, b
w.1, ub
w.2, ub
f.1, ub
w.1, ub
w.2, ub;
f.2, ub
Persoonia
falcata
Terminalia
latipes
Vitex
glabrata
Uraria
lagopodiodies
Unidentified
O200ub
w.1, ub;
f.4, ub
w.2, ub
w.13, ub; w.13, ub; w.9, ub; w.5, ub;
f.19, ub
f.92, ub
f.39, ub f.28, ub
w.2, b; f.2, ub
w.1, ub;
f.4, ub
w.20, ub; w.4, ub; f.4, ub
f.35, ub
f.17, ub
f.3, ub
w.8, ub; w.16, ub;
w1, b;
f.18, ub
f.29, ub
f.3, ub
w.1, ub
w.1, ub
J. Atchison et al. / Journal of Archaeological Science 32 (2005) 167–181
Allotropis
semialata
Alysicarpus sp.
Asteraceae
Buchanania
obovata
Cajanus sp.
Crotalaria
crispata
Desmodium sp.
J3
w.1, b;
f.1, ub
w.1, ub;
f.2, ub
w.51, ub
25, ub
42, ub
61, ub
87, ub
51, ub
82, ub
75, ub
O100, ub 46, ub
71, ub
4 ub; 4 b w.3, ub;
f.22, ub
f.17, ub
w.2, ub;
f.57, ub
171
J1-9 (Jinmium surface sites 1–9). M10–16 (Marralam surface sites 10–16).
w denotes whole seeds; f denotes fragmented seeds; ub denotes unburnt seeds; b denotes burnt seeds.
80, ub
172
s
ra
gl
ab
Vi
te
x
Te
ta
la
tip
e
ta
oo
rs
Pe
lia
ni
a
an
an
ch
Bu
rm
in
a
ob
ia
an
og
op
H
et
er
fa
lc
a
ov
nt
or
co
in
ic
a
ja
v
us
er
Cy
p
Depth
cm
at
a
tu
s
J. Atchison et al. / Journal of Archaeological Science 32 (2005) 167–181
0
Unit 7
Unit 6
10
20
Unit 5
30
Unit 4
40
50
Unit 3
60
70
Unit 2
80
90
100
110
Unit 1
120
130
140
No. of whole and
fragmented seeds
per litre x 10-1
0
0.5
0
0.5
0
0.5
1
1.5
0
0.5
0
0.5
0
0.5
Fig. 3. C1/I Pit, Jinmium, Species Assemblage O4 mm. Species assemblage from pit C1/I from the O4 mm fraction. Numbers of whole and
fragmented seeds are measured per litre. No seed material was dated from this pit.
cliff within a sheltered gorge. Two trenches were excavated in the deepest area with a sandy floor: PP1 was
excavated close to the drip line, and PP2 approximately
5 m away close to the back wall. Both pits are
characterised by loosely packed sandy sediment with
a high charcoal content. The maximum depth was 90 cm
in PP1, with cultural material in the upper 60 cm. The
stratigraphy of the pit was characterised by predominantly loosely packed sandy sediment with a high
charcoal content. Both pits contained a variety of
cultural materials including flaked stone, grinding
stones, stone points, bone, ochre, shell and glass. PP2
sediment is very dry and loose, and likely to be disturbed
and mixed by kangaroo hollows.
Fruit tree seeds dominate all units of the PP1 seed
assemblage (Figs. 9, 10). B. obovata, T. latipes and
Terminalia ferdinandiana form the major components of
both the O4 mm and 2–4 mm fractions. A major peak
of P. falcata and B. obovata seed fragments was recovered from spit 6B (33–43 cm). Associated stratigraphic evidence indicates that this is a pit dug into spit
6. All of the seeds more than 15 cm below the surface
were burnt. The species assemblage in pit PP2 is also
dominated by fruit tree species.
5. Discussion
5.1. Distinguishing seeds deposited by cultural and
non-cultural means
The modern seed assemblages are characterised first
by the high density of grass and other small seed species,
and second by spatial concentrations of mostly unburnt
fruit tree seeds near the parent tree. Small seeds such as
those of Panicum sp. or Y. majuscula are fairly light and
173
es
tip
ov
ed
la
ifi
ia
nt
al
de
ni
U
Te
Bu
in
an
ch
m
es
D
rm
iu
od
an
m
ia
sp
ob
.
de
tri
a
gi
ste
no
Xe
et
H
at
at
nt
or
nt
co
an
og
op
er
14
ag C (A
es
(y MS
rB )
P)
Depth
cm
a
a
tu
s
J. Atchison et al. / Journal of Archaeological Science 32 (2005) 167–181
0
Unit 7
Unit 6
10
Unit 5
20
30
1050±60 (Wk7691)
1210±60 (Wk7692)
Unit 4
40
50
Unit 3
60
Unit 2
70
80
1060±70 (Wk7693)
90
100
110
Unit 1
120
130
3400±60 (Wk7694)
140
150
No. of whole and
fragmented seeds
per litre x 10-1
0
0.1 0.2 0.3 0.4
0
0.1
0
0.1
0
0.1
0
0.1 0.2
0
0.1
Fig. 4. C1/III Pit, Jinmium, Species Assemblage O4 mm. Species assemblage from pit C1/III from the O4 mm fraction. Numbers of whole and
fragmented seeds are measured per litre. 14C (AMS) dates are sourced from Persoonia falcata seed material dated at the Waikato dating laboratory
(these results are presented in Table 2).
Table 2
Summary of dates obtained in this study
Site
Depth (cm)
Spit
Material
Technique
Age
Lab. code
C1/III
C1/III
C1/III
C1/III
G1/I
G1/I
G1/I
G1/I
G1/I
G1/I
G1/I
G1/I
PP1
PP1
PP1
33–39
33–39
80–84
127–137
22.4–27.6
43–48.75
43–48.75
118.43–128.14
96.83–107.75
96.83–107.75
128.14–135.33
107.75–118.43
33.25–43.25
33.25–43.25
33.25–43.25
10B
10B
18A
25A
6
10
10
20
18A
18A
21A
19
6B
6
6
seed fragment
seed fragment
seed fragment
seed fragment
charcoal
seed fragment
seed fragment
seed fragment
seed fragment
charcoal
charcoal
pit sediment
seed fragment
charcoal
seed fragment
14
1050G60
1210G60
1060G70
3400G60
200G60
2620G60
2970G70
contaminated
1780G90
6170G200
3202G59
8400G100
3330G60
3500G100
3540G70
Wk7691
Wk7692
Wk7693
Wk7694
Wk7301
Wk7696
Wk7695
C AMS
14
C AMS
14
C AMS
14
C AMS
conventional
14
C AMS
14
C AMS
14
C AMS
14
C AMS
conventional
conventional
TL
14
C AMS
conventional
14
C AMS
14
C
14
C
14
C
14
C
Wk7895
Wk7302
Wk7303
W2459
Wk7699
Wk7700
Wk7698
174
nt
ifi
e
U
ni
de
Vi
te
x
gl
ab
ra
d
ta
la
tip
e
lia
fa
lc
a
ni
a
oo
Te
Pe
rs
rm
in
a
ov
ow
ni
i
ob
br
ia
ve
Pa
Bu
Ac
tta
an
ch
ia
ac
an
sp
.
gi
a
ste
no
Xe
ta
at
a
ta
ta
en
tri
d
.
sp
m
ni
cu
Pa
s
m
el
lip
tic
u
yl
um
Er
yt
hr
op
H
et
er
ox
og
.
G
ly
c
in
e
sp
ia
ta
la
r
Cr
o
Depth
cm
0
an
sp
.
co
nt
or
tu
s
J. Atchison et al. / Journal of Archaeological Science 32 (2005) 167–181
Unit 7
Unit 6
10
Unit 5
20
30
Unit 4
40
50
Unit 3
60
70
Unit 2
80
90
100
110
Unit 1
120
130
140
No. of whole and
fragmented seeds 0 8 16 0 8 16 0
per litre
x 10-3
x 10-3
4
0
0.4 0
8
16
x 10-3
0
8
x 10-2
0
8
x 10-2
16 0
20 0
x 10-2
8
x 10-2
16 0
2 0
0.8 1.6 0
4
8 0
0.4
x 10-2
Fig. 5. C1/I Pit, Jinmium, Species Assemblage 2–4 mm. Species assemblage from pit C1/I from the 2–4 mm fraction. Numbers of whole and
fragmented seeds are measured per litre. No seed material was dated from this pit.
may be moved or blown around a large area quite
readily. Their widespread distribution most likely reflects this process. Fruit seeds may also be moved by
birds, but in general, the main component of seed
assemblages underneath a large fruit tree are seeds of the
same species. Large and or heavy seeds are not well
distributed away from larger fruit trees.
The upper Units of each of the rockshelter excavation
pits contain a significant component of grass, other
small seed species and/or fruit tree seeds, most of which
are unburnt. The high density of grass seeds and low
density of unburnt fruit seeds (similar to the off site
assemblages) support the idea that the upper Units as
defined are essentially non-cultural assemblages. The
whole unburnt seeds in the upper layers of CI/I–III were
probably introduced to the site through chance distribution by birds, since this site is not overhung by fruit
trees. Fruit seeds in the upper layers were always found
in relatively low densities. There are no perches or
crevices underneath the C1 rockshelter and it would
seem a fairly unsuitable site for animal seed caching. The
topography of the sand sheet surrounding the shelter is
also quite flat so it is unlikely that seeds would collect
there by rolling or water deposition.
In contrast, the high densities of burnt fragments of
B. obovata and P. falcata seeds in the lower levels of all
sites are considered to be cultural (Fig. 11). Processing
of P. falcata and B. obovata has been recorded in a
number of north Australian ethnographic and archaeological studies. Smith and Kalotas [31] and Kenneally
et al. [22] describe the processing sequence of P. falcata
by the Bardi people from the western Kimberley. In this
case ‘‘the edible fruit (is) usually collected from the
ground and eaten raw when ripe [yellow]; (or the) edible
seed pounded, (and) mixed with water to make a black
custard’’ ([22]: 171). B. obovata also has extremely hard
seeds and we have recorded a similar processing
sequence of B. obovata by senior members of the
Marralam community. Both species are known as wet
season fruit sources to local Aboriginal senior women.
The fruit is picked or collected when ripe, mashed into
a paste with a large stone and often mixed with sugar
175
s
nt
ifi
e
d
ta
ra
gl
ab
lia
Te
Vi
te
x
rm
in
a
oo
rs
Pe
U
ni
de
la
tip
e
ta
fa
lc
a
ni
a
br
Pa
ve
tta
ox
an
an
yt
hr
Er
ch
Bu
el
lip
tic
u
ob
yl
um
ia
e
ea
M
im
os
ac
ex
ow
ni
i
at
a
ov
la
ta
sti
pu
en
gi
a
ste
m
od
no
Xe
D
es
Ca
ly
tri
x
sp
iu
m
an
og
op
H
et
er
14
ag C (
es A
(y MS
rB )
P)
Depth
cm
0
tri
d
.
co
nt
or
ta
ta
tu
s
m
J. Atchison et al. / Journal of Archaeological Science 32 (2005) 167–181
Unit 7
Unit 6
10
Unit 5
20
30
40
1050±60(Wk7691)
1210±60(Wk7692)
Unit 4
50
Unit 3
60
Unit 2
70
80
1060±70(Wk7693)
90
100
110
Unit 1
120
130
3400±60(Wk7694)
140
150
No. of whole and
fragmented seeds per
litre x10-1
0
0.2
0.4
0.6
0
0
0.2
0
0.2
0
0.2
0.4 0
0
0
0
0.2
0.4
0
0
0
0.2
Fig. 6. C1/III Pit, Jinmium, Species Assemblage 2–4 mm. Species assemblage from pit C1/III from the 2–4 mm fraction. Numbers of whole and
fragmented seeds are measured per litre. 14C (AMS) dates are sourced from Persoonia falcata seed material dated at the Waikato dating laboratory
(these results are presented in Table 2).
(Fig. 12). The mash is consumed raw and any seed spat
out (for full details see [16]). Processing such as
described above would account for the fragmentation.
With regard to the seeds being burnt, Smith and
Kalotas [31] report that because P. falcata seeds are so
hard, they are pounded and warmed in hot ash.
Crawford ([8]: 46–47) noted that P. falcata fruits at
Kalumburu in the west Kimberley were sun-dried,
cooked in ashes, hammered and then stored in paperbark. The high density of P. falcata seeds in the inferred
cooking pit at Punipunil is consistent with this process.
We have found no records of B. obovata having been
burnt in this way but the burnt nature and consistent
size (dominating the 2–4 mm fractions) of both species
in the seed assemblages suggest they have both been
processed in a manner similar to that recorded above. It
is possible that the cooking process makes the tough
seeds easier to open, as with other hard seeds (e.g.
Macadamia sp.). While it is also possible that the fruits
were first smashed and the hard seed coats discarded in
the fire (a procedure used to discourage ants in a camp
site, Wightman, pers. comm., 1998), it is difficult to
determine the exact order of processing events from the
remains themselves. What is clear is that each of the
three rockshelters has been an important focus for fruit
processing. The implications of potential storage provided by dried fruit cakes are considerable, and could
relate to situations of resource scarcity, travel and/or
ceremonial activity.
5.2. Chronology, taphonomy and the onset of fruit
seed processing
The onset of sustained fruit seed processing occurred
about 3500 BP at Punipunil I, with comparable dates for
bulk charcoal and P. falcata seeds in Unit 2. The lowest
seeds occur slightly above the lowest cultural material in
this site. There is flaked stone (O4 mm) in spit 7 (unit 1)
of PP1 and the lowest seeds are in Spit 6 and 6B (Unit 2)
of PP1.
176
s
nt
ifi
e
d
la
tip
e
Te
U
ni
de
ni
a
oo
rs
lia
rn
ve
ia
Pe
O
we
n
rm
in
a
ic
o
fa
lc
a
ta
sa
at
a
ov
ob
ia
an
an
Bu
Depth
cm
ch
14
14 C
(c
TL C
( on
ag AM ven
es S) tio
( y a n na
rB d
l),
P)
J. Atchison et al. / Journal of Archaeological Science 32 (2005) 167–181
Unit 3
10
20
200±60 (Wk7301)
Unit 2
30
40
50
2620±60 (Wk7696)
2970±70 (Wk7695)
60
70
80
90
Unit 1
100
6170±200 (Wk7302)
1780±90 (Wk7895)
110
8400±100 (W2459)
120
130
3202±59 (Wk7303)
No. of whole and
fragmented seeds
per litre x 10-1
0
0.1 0.2 0.3
0
0.1
0
0.1
0
0.1
0
0.1 0.2 0.3
Fig. 7. G1/I Pit, Granilpi, Species Assemblage O4 mm. Species assemblage from pit G1/I from the O4 mm fraction. Numbers of whole and
fragmented seeds are measured per litre. Conventional 14C (italics underlined) and 14C (AMS) are sourced from charcoal and Persoonia falcata seed
material, respectively, dated at the Waikato dating laboratory. The TL date (bold) is sourced from sand dated at the University of Wollongong TL
laboratory (these results are presented in Table 2).
At Granilpi abundant seeds at the base of the
charcoal-rich Unit 2 suggest an onset at about 3000 BP.
Occasional seed fragments occur in the lower sandy Unit
1, but we cannot rule out down-profile movement of
seeds and small charcoal fragments. The latter explanation would account for the seed date of 1780 G 90
(Wk7895) at a depth with a bulk charcoal date of
6170 G 200 (Wk7302). High frequency of stone artefactual material to the base of the deposit indicates that
occupation of the site occurred well before the onset of
seed processing.
At Jinmium the oldest seed date is 3400 G 60
(Wk7694) at 132 cm, with the most intensive deposits
in younger layers. Seeds from the lower units of CI/III
were noticeably weathered. This and the more gradual
decline in seeds with depth (compared to the more
abrupt changes at Granilpi) indicate that preservational
factors rather than actual onset of the activity are
influencing the record more at Jinmium. The chronology
provided by the seeds does not, on its own, suggest any
significant movement of seed material within the C1/I, II
or III trenches, although movement of carbon material
and mixing of sediment below 1 m depth are indicated by
the range of radiocarbon dates (conventional, AMS, and
elemental carbon) and luminescence age estimates (TL
and OSL; multiple aliquots, single aliquots and single
grain) discussed by Roberts et al. [27]. The oldest seed
date is quite consistent with the seed chronology as
a whole, although its depth well below any stone points –
that here and elsewhere in the study area are consistently
younger than about 3500 years – lends support to
arguments for significant movement and/or mixing in
177
s
U
ni
de
nt
ifi
e
ra
gl
ab
Vi
te
x
Te
d
ta
la
tip
e
lia
rm
in
a
oo
rs
Pe
Pa
Bu
ve
ch
tta
an
ni
a
an
br
ia
fa
lc
a
ob
ow
ni
i
ov
ta
at
a
tu
s
nt
or
co
an
og
op
H
et
er
14
Depth
cm
14 C
TL C (con
ag (AM ven
es S tio
( y ) a na
r B nd l)
,
P)
J. Atchison et al. / Journal of Archaeological Science 32 (2005) 167–181
Unit 3
10
20
200±60 (Wk7301)
30
Unit 2
40
50
2620±60 (Wk7696)
2970±70 (WK7695)
60
70
80
90
Unit 1
100
110
6170±200 (Wk7302)
1780±90 (Wk7895)
8400±100 (W2459)
120
130
3202±59 (Wk7303)
No. of whole and
fragmented seeds
per litre
0
2
4
0
0.4
-2
x 10
0.8
1.2
0
4
8
-3
x 10
12
16
0
1
2
0
10
-2
x 10
0
2
0
0.4
0.8
1.2
1.6
-2
x 10
Fig. 8. G1/I Pit, Granilpi, Species Assemblage 2–4 mm. Species assemblage from pit G1/I from the 2–4 mm fraction. Numbers of whole and
fragmented seeds are measured per litre. Conventional 14C (italics underlined) and 14C (AMS) are sourced from charcoal and Persoonia falcata seed
material, respectively, dated at the Waikato dating laboratory. The TL date (bold) is sourced from sand dated at the University of Wollongong TL
laboratory (these results are presented in Table 2).
Unit 1 sediments. On the other hand, the date is
consistent with a Holocene chronology for human
occupation in Unit 1 (contra [11]), as proposed by
Roberts et al. [27].
5.3. Spatial and temporal trends in fruit
seed processing
Any exploitation of fruit in the region would have
been a highly seasonal activity, since the timing of fruit
availability is locked in to the timing of the wet season,
which has considerable monthly variation. Both species
flower between July and October and fruit between
October and December, but fruiting is also recorded in
the literature through January and February [7,29].
Occupation of the rock shelters to process fruit is thus
most likely to have taken place sometime between
October and February, coinciding with the onset and
development of the monsoon.
Within individual sites, particularly intense phases of
fruit seed deposition occur, for example in Jinmium pit
III about one thousand years ago, with overlapping seed
dates from samples more than 40 cm apart. However,
the variability evident in the three Jinmium pits and two
Punipunil pits suggests that these reflect spatial variability within sites (cf. [33]) rather than any sort of
broader trend.
At all sites fruit seed processing was one of a number of activities being undertaken. For example, the
Jinmium pit III phase also has high densities of ochre
and stone points. At Granilpi there is also a correlation
of seed processing with artefact densities including stone
point production in the lower half of Unit 2. Initial
evidence for exploitation of seeds seems to precede any
direct evidence of stone point production [1].
One trend visible at all three sites is the decline in
processed B. obovata and P. falcata seed fragments
towards the top of each site. Above the decline in
178
nd
lia
nt
ifi
e
d
la
tip
e
s
di
na
fe
r
lia
U
ni
de
Te
Te
rm
in
a
rm
in
a
rs
Pe
Bu
C
ch
oo
an
an
ni
a
ia
ob
fa
lc
a
ov
ta
at
a
(
(A stan
M da
S) r d
ag ) a
e s nd
(y
rB
P)
St
em
od
ia
ly
th
rif
ol
ia
1
14 C4
Depth
cm
0
ia
n
a
J. Atchison et al. / Journal of Archaeological Science 32 (2005) 167–181
Unit 5
Unit 4
10
20
Unit 3
30
3330±60 (Wk7699)
40
Unit 2
3500±100 (Wk7700)
3540±70 (Wk7698)
50
Unit 1
60
70
sterile
80
No. of whole and
4
fragmented seeds 0
-3
per litre
x 10
8
12
16
0
2
4
6
8
-2
x 10
0
10
x 10
-2
20
0
4
-3
x 10
8
12
16
0
0.2
0.4
0
2
-2
x 10
Fig. 9. PP1 Pit, Punipunil, Species Assemblage O4 mm. Species assemblage from pit PP1 from the O4 mm fraction. Numbers of whole and
fragmented seeds are measured per litre. Conventional 14C (italics underlined) date is based on charcoal and 14C (AMS) on Persoonia falcata seed
material, respectively, dated at the Waikato dating laboratory (these results are presented in Table 2).
processed fragments, both species only occur as whole
and unburnt seeds, suggesting that they are non-cultural
in origin. The upper units of each pit also contain a
significant component of grass and other small seed
species and or fruit tree seeds, most of which are
unburnt, similar to the off site assemblages. Available
dates and occurrence in these levels of post contact
archaeological materials such as glass indicate a date for
this phase within the last 150 years.
This phase is not simply one of site abandonment,
but represents a reorientation in activities associated
with the arrival of pastoralism. This includes intensification of point production at a number of sites (cf.
[1,15]). Even so, since we know from ethnographic evidence that wet season occupation of many traditional
sites was maintained during ‘holiday times’ from station
work [26], it could be expected that fruit seed processing
would be part of this occupation. Significantly however,
during detailed mapping and observation of each
excavated site [3], no viable or adult P. falcata were
recorded in a minimum 1 km radius. The species is
considered marginal at Jinmium and Granilpi and nonexistent at Punipunil, with a small but regenerating
population at Marralam. The most likely explanation
is the shift to more intense late dry season fires as
Aboriginal patterns of burning throughout the dry
season were disrupted with colonisation. Intense fires
have been identified elsewhere by Aboriginal people as
impacting negatively on P. falcata flowering and fruiting
[14,34]. Further research is needed on this issue.
6. Conclusion
The picture of plant use presented in this paper is
an artefact of fortuitous taphonomic processes. Here
preservation is a direct result of Aboriginal people
burning and carbonising the particularly dense hard
seeds as part of the fruit processing technique, a process
179
s
d
nt
ifi
e
rm
in
a
Te
U
ni
de
lia
fa
lc
a
ni
a
oo
rs
Pe
ch
Bu
la
tip
e
ta
at
a
ov
ob
ia
an
an
sp
.
ia
ac
Ac
C
1
14 C4
Depth
cm
0
(
(A stan
M da
S) r d
ag ) a
e s nd
(y
rB
P)
J. Atchison et al. / Journal of Archaeological Science 32 (2005) 167–181
Unit 5
Unit 4
10
20
Unit 3
30
3330±60 (Wk7699)
40
Unit 2
3500±100 (Wk7700)
3540±70 (Wk7698)
50
Unit 1
60
70
sterile
80
No. of whole and
fragmented seeds
per litre
0
1
2
3
4
0 0.5 1 1.5
0
1
2
-2
x 10
3
4
5
6
7
0 0.1 0.2 0.3 0.4
0
2
4
6
8
10 12 14
x 10-2
Fig. 10. PP1 Pit, Punipunil, Species Assemblage 2–4 mm. Species assemblage from pit PP1 from the 2–4 mm fraction. Numbers of whole and
fragmented seeds are measured per litre. Conventional 14C (italics underlined) date is based on charcoal and 14C (AMS) on Persoonia falcata seed
material, respectively, dated at the Waikato dating laboratory (these results are presented in Table 2).
undertaken in a unique microenvironment, that is, small
isolated sandstone rock shelters in the surrounding
sandy plain.
While P. falcata and B. obovata were both found in
high densities in each of the archaeological excavations,
it is not suggested that these were the only plants processed. In fact, the morphology of the remains that were
found suggests that it was only those species with hard,
charred seed coats that survived decomposition. Fruits
with more fragile seeds that were consumed raw,
including V. glabrata, T. ferdinandiana and T. latipes,
are strongly represented in the upper archaeological
levels, but were presumably also eaten during earlier
times for which the evidence has not survived as well.
The archaeological record of even less well preserved
plants such as yams and other tubers is being pursued in
this project via studies into starch residues on stone
tools.
Burning of the seeds through processing has enabled
the cultural and non-cultural components to be quite
clearly distinguished and allows us one window into
long term and more recent patterns of Aboriginal plant
use. This picture depicts seasonal and fluctuating occupation of the small sandstone rockshelters while fruit
was available over the late Holocene. Nevertheless the
ethnographic record allows us to give considerable life
to the archaeological window. We know that, although
short-lived, the wet season fruits provide abundant
foods highly valued by both adults and children for
their taste and ease of access. We can envisage large
scale processing of the fruits by groups of women to
facilitate longer term storage. The implications of such
storage for travel, ceremonial gatherings and intergroup exchange are being considered in ongoing work
that compares the fruit seed window with other
archaeological windows such as those of stone artefacts
and ochre.
In contrast however, evidence of fruit processing
ceases with the arrival of pastoralism and changed fire
regimes, suggesting a significant shift in peoples’ utilization of these rocky outcrops, and potentially also a
change in the ecology of these environments.
180
J. Atchison et al. / Journal of Archaeological Science 32 (2005) 167–181
Fig. 11. Percentage of burnt and unburnt seed material in each excavation pit.
Acknowledgements
We are indebted to Biddy Simon and Polly Wandanga
of the Marralam community for collaboration over
many years. Assistance with excavation and plant
recording was provided by Maurice Simon, May Melpi,
Paul Melpi, Iza Pretlove, Eileen Huddleston, and Ken
Mulvaney. Funding was provided by the Australian
Research Council and the University of Wollongong
(Research Centre for Landscape Change).
References
Fig. 12. Ripe Buchanania obovata being pounded into a mash.
[1] K. Akerman, R. Fullagar, A. van Gijn, Weapons and Wunan:
production, function and exchange of spear points from the
Kimberley, northwestern Australia, Australian Aboriginal Studies
2002 (1) (2002) 13–42.
[2] R. Allen, An investigation of past Aboriginal plant use using
macroscopic plant reproductive remains from Jinmium and
Punipunil rock-shelters, Northern Territory, Honours thesis,
University of Wollongong, 1997.
[3] J. Atchison, Continuity and change: a late Holocene and post
contact history of Aboriginal environmental interaction and
vegetation process from the Keep River region, Northern
Territory, PhD thesis, University of Wollongong, 2000.
[4] C.R. Cartwright, Grapes or raisins? An early Bronze Age larder
under the microscope, Antiquity 77 (2003) 345.
[5] A. Clarke, A preliminary archaeobotanical analysis of the
Anbangbang I site, in: R. Jones (Ed.), Archaeological Research
in Kakadu National Park, Australian National Parks and Wildlife
Service, Canberra, 1985, pp. 77–96.
J. Atchison et al. / Journal of Archaeological Science 32 (2005) 167–181
[6] A. Clarke, Macroscopic plant remains, in: W. Beck, A. Clarke,
L. Head (Eds.), Plants in Australian Archaeology, University of
Queensland, St. Lucia, 1989, pp. 54–89.
[7] M. Clark, S. Traynor, Plants of the Tropical Woodland, Conservation Commission of the Northern Territory, Darwin, 1987.
[8] I.M. Crawford, Traditional Aboriginal plant resources in the
Kalumburu area: aspects in ethno-economics, Records of the
Western Australian Museum 15 (1982).
[9] I. Farrington, J. Urry, Food and the early history of cultivation,
Journal of Ethnobiology 5 (1985) 143–157.
[10] H. Fujiwara, R. Jones, S. Brockwell, Plant opals (phytoliths) in
Kakadu archaeological sites, in: R. Jones (Ed.), Archaeological
Research in Kakadu National Park, Australian National Parks
and Wildlife Service, Canberra, 1985, pp. 155–164.
[11] R.L.K. Fullagar, D.M. Price, L.M. Head, Early human occupation of northern Australia: archaeology and thermoluminescence
dating of Jinmium rock-shelter. Northern Territory, Antiquity
70 (1996) 751–773.
[12] D.R. Harris, G.C. Hillman (Eds.), Foraging and Farming: the
Evolution of Plant Exploitation, Unwin Hyman, London, 1989.
[13] C.A. Hastorf, The cultural life of early domestic plant use,
Antiquity 72 (1998) 773–782.
[14] C.D. Haynes, Use and impact of fire, in: C.D. Haynes,
M.G. Ridpath, M.A.J. Williams (Eds.), Monsoonal Australia
Landscape, Ecology and Man in the Northern Lowlands, A.A.
Balkema, Rotterdam, 1991, pp. 61–71.
[15] L. Head, R. Fullagar, Hunter–gatherer archaeology and pastoral
contact: perspectives from the northwest Northern Territory,
Australia, World Archaeology 28 (1997) 418–428.
[16] L. Head, J. Atchison, R. Fullagar, Country and garden.
Ethnobotany, archaeobotany and Aboriginal landscapes near
the Keep River, northwestern Australia, Journal of Social
Archaeology 2 (2002) 173–196.
[17] N. Horsfall, Holocene occupation of the tropical rainforests of
north Queensland, in: P. Veth, P. Hiscock (Eds.), Archaeology of
Northern Australia, University of Queensland, St. Lucia, 1996,
pp. 175–190.
[18] R.A. Hynes, A.K. Chase, Plants, sites and domiculture: Aboriginal influence on plant communities in Cape York, Archaeology
in Oceania 17 (1982) 38–50.
[19] S. Johannessen, C.A. Hastorf (Eds.), Corn and Culture in the
Prehistoric New World, Westview Press, Boulder, CO, 1994.
[20] R. Jones, Archaeological Research in Kakadu National Park,
Australian National University, Canberra, 1985.
[21] R. Jones, B. Meehan, Plant foods of the Gidjingali: ethnographic
and archaeological perspectives from northern Australia on tuber
and seed exploitation, in: D.R. Harris, G.C. Hillman (Eds.),
Foraging and Farming: the Evolution of Plant Exploitation,
Unwin Hyman, London, 1989.
181
[22] K.F. Kenneally, D.C. Edinger, T. Willing, Broome and Beyond:
Plants and People of the Dampier Peninsular, Kimberley,
Western Australia, Department of Conservation and Land
Management, Como, 1996.
[23] K. McConnell, Palaeoethnobotanical remains of Carpenter’s Gap
Site 1, the Kimberleys, Western Australia, Masters thesis, The
Australian National University, 1996.
[24] K. McConnell, S. O’Connor, 40,000 Year record of food plants in
the Southern Kimberley Ranges, Western Australia, Australian
Archaeology 45 (1997) 20–31.
[25] M. Morwood, Art and Stone: towards a prehistory of central
western Queensland, PhD thesis, Australian National University,
1979.
[26] K. Mulvaney, What to do on a rainy day. Reminiscences of
Miriuwung and Gajerrong artists, Rock Art Research
13 (1996) 3–20.
[27] R. Roberts, M. Bird, J. Olley, R. Galbraith, E. Lawson,
G. Laslett, H. Yoshida, R. Jones, R. Fullagar, G. Jacobsen,
Q. Hua, Optical and radiocarbon dating at Jinmium rock
shelter in northern Australia, Nature 393 (1998) 358–
362.
[28] J. Russell-Smith, D. Lucas, M. Gapindi, B. Gunbunuka,
N. Kapirigi, G. Namingum, K. Lucas, P. Giuliani,
G. Chaloupka, Aboriginal resource utilization and fire management practice in western Arnhem Land, monsoonal Northern
Australia: notes for prehistory, lessons for the future, Human
Ecology 25 (2) (1997) 159–195.
[29] B.L. Rye, B.L. Koch, A.J.G. Wilson, in: J.R. Wheeler (Ed.), Flora
of the Kimberley Region, Western Australian Herbarium,
Department of Conservation and Land Management, Perth,
1992.
[30] C. Schrire, The Alligator Rivers: Prehistory and Ecology in
Western Arnhem Land, Terra Australis, Department of
Prehistory, Australian National University, Canberra, 1982.
[31] M. Smith, A.C. Kalotas, Bardi plants: an annotated list of plants
and their use by the Bardi Aborigines of Dampier land, Northwestern Australia, Records of the West Australian Museum
12 (1985) 317–359.
[32] P. Tacon, R. Fullagar, S. Ouzman, K. Mulvaney, Cupule
engravings from Jinmium-Granilpi (northern Australia) and
beyond: exploration of a widespread and enigmatic class of rock
markings, Antiquity 71 (1997) 942–965.
[33] R. Theunissen, J. Balme, W. Beck, Headroom and human
trampling: cave ceiling-height determines the spatial patterning
of stone artefacts at Petzkes Cave, northern New South Wales,
Antiquity 72 (1998) 80–90.
[34] R.J. Williams, Fire and floral phenology in a humid tropical
savanna at Kapalga, Kakadu National Park, northern Australia,
Proceedings, Bushfire ’97, 1997.