ArticlePDF Available

Revisiting Mwulu’s Cave: new insights into the Middle Stone Age in the southern African savanna biome

  • GEM (Guías de Espeleología y Montaña)

Abstract and Figures

In this paper, we present a revised stratigraphy and results of preliminary analyses of the archaeological material from Mwulu’s Cave. This arises from two excavation campaigns conducted in 2017, 71 years after the site was initially investigated by P.V. Tobias. This cave, located in Limpopo Province (South Africa), preserves one of the few known Middle Stone Age sequences in the northeastern part of the country. Here, we revisit the stratigraphic sequence of the site and provide new analyses of sediments, palynomorphs, phytoliths, ochre and lithics. The renewed excavations and reappraisal of the archaeological material from Mwulu’s Cave form part of a larger research project exploring Middle Stone Age variability in the northeastern part of South Africa, with a specific focus on the so-called Pietersburg industries.
This content is subject to copyright. Terms and conditions apply.
Revisiting Mwulus Cave: new insights into the Middle Stone Age
in the southern African savanna biome
P. de la Peña
&A. Val
&D. J. Stratford
&F. Colino
&I. Esteban
&J. M. Fitchett
&T. Hodgskiss
&J. Matembo
R. Moll
Received: 20 September 2018 / Accepted: 31 October 2018
#Springer-Verlag GmbH Germany, part of Springer Nature 2018
In this paper, we present a revised stratigraphy and results of preliminary analyses of the archaeological material from Mwulus
Cave. This arises from two excavation campaigns conducted in 2017, 71 years after the site was initially investigated by P.V.
Tobias. This cave, located in Limpopo Province (South Africa), preserves one of the few known Middle Stone Age sequences in
the northeastern part of the country. Here, we revisit the stratigraphic sequence of the site and provide new analyses of sediments,
palynomorphs, phytoliths, ochre and lithics. The renewed excavations and reappraisal of the archaeological material from
Mwulus Cave form part of a larger research project exploring Middle Stone Age variability in the northeastern part of South
Africa, with a specific focus on the so-called Pietersburg industries.
Keywords Stratigraphy .Sedimentology .Pollen .Phytoliths .Ochre .Lithics .Pietersburg industry
Mwulus cave: a new project
The literature for the Middle Stone Age (MSA) archaeology
of southern Africa is far from being clear. The MSA, a term
somewhat ambiguously defined in the late 1920s (Goodwin
and Van Riet Lowe 1929), has had supporters and detractors
since its conception. Nonetheless, it has been retained as a
useful framework by most researchers. The MSA in southern
Africa has included, from its initial definition until now, a
variety of industries and regional expressions, which include
amongst many others: the Still Bay, Proto-Still Bay,
Pietersburg, Bambatan, Howiesons Poort, Hagenstad,
Mazelspoort, Magosian, Pre-Still Bay, Post-Howiesons
Poort, and the Sibudan. Some of these industries, such as the
Still Bay, were vehemently rejected in the second half of the
twentieth century (Sampson 1972,1974), but now have strong
defenders (Wadley 2008; Henshilwood 2012; Conard and
Porraz 2015). This contrasts with others, notably the
Howiesons Poort, which had a complex stratigraphic position
and was considered to represent industries mixed with other
assemblages (Magosian) and is now again in an ambiguous
chrono-stratigraphical position and thus the focus of new ac-
ademic debate (e.g. Jacobs and Roberts 2015; Tribolo et al.
2009,2013), although no one would dispute its idiosyncrasy
anymore. Finally, some initially weakly defined variations,
such as the Pietersburg, went through a golden agebefore
falling into oblivion. Nonetheless, this term has recently been
resumed (Lombard et al. 2012; Porraz et al. 2015,2018). This
paper contributes directly to the latter: the lithic material re-
trieved from Mwulus Cave was indeed fully attributed to this
Electronic supplementary material The online version of this article
( contains supplementary
material, which is available to authorized users.
*P. de la Peña
Evolutionary Studies Institute, University of the Witwatersrand,
Johannesburg, South Africa
School of Geography, Archaeology and Environmental Studies,
University of the Witwatersrand, Johannesburg, South Africa
Department of Early Prehistory and Quaternary Ecology, University
of Tübingen, Schloss Hohentübingen, 72070 Tübingen, Germany
GEM (Guías de Espeleología y Montaña), Casilla del Mortero,
Torremocha del Jarama, 28189 Madrid, Spain
African Centre for Coastal Palaeoscience, Nelson Mandela
University, Port Elizabeth, Eastern Cape 6031, South Africa
Origins Centre, University of the Witwatersrand,
Johannesburg, South Africa
Archaeological and Anthropological Sciences
industry (Tobias 1949). Different sites in Limpopo Province
were also attributed to the Pietersburg by Mason (1957): Cave
of Hearths, Olieboomspoort, Rufus Cave, Kalkbank,
Rooirand, Border Cave, Bushman Rock Shelter and
Koedoesrand. Mason (1957) paid special attention to this in-
dustry and published some of the first statistical comparisons
on southern African stone tools. Therefore, the Pietersburg
was arguably, for some time, one of the most accurately de-
scribed industries in Africa.
When Mwulus Cave was excavated by Tobias in 1947,
this site was only the second sequence found in a cave with
a sealed depositattributed to the Pietersburg, following
Border Cave (Cooke et al. 1945). It seems ironic that Tobias
mentions several problems relating to the definition of the
Pietersburg at the beginning of his paper on Mwulus Cave:
Was it of long duration or short? Was it a single cultural
manifestation, or did it show a range of variation within itself?
The answers were necessary in order to view the Pietersburg
in proper perspective(Tobias 1949: p. 2). Tobias thought that
the recentat that timeexcavation of Border Cave by
Malan, Cooke and Wells (1945), a newly excavated sealed
deposit, would solve all questions surrounding the
Pietersburg. In a similar manner, Mwulus Cave was viewed
as a good candidate to solve those issues. Seventy-one years
later, however, we still have analogous questions. In the two
publications of Tobias (1949,1954), a stratigraphy and a gen-
eral description of the archaeological materials were provided,
and a hypothesis on the accumulation of sediment in the cave
was proposed (Tobias 1954).
In 2017, a new project at Mwulus Cave was initiated with
the following objectives: to provide a revised stratigraphy, as
well as a chronological and palaeoenvironmental framework,
and to re-study, technologically, the archaeological material
from Tobiass and the new collections. The ultimate goal
was to include all of these data in a renewed discussion on
the MSA lithic variability in the northeastern part of southern
Africa. In trying to understand the reasons behind this vari-
ability, one can propose hypotheses on these alleged periods
of change or stasis. The latter will require further work, since
we are still in a descriptive phase.
Exploring the variability of the MSA in that region was
deemed an important task to accomplish for several reasons.
First, although at some stage during the twentieth century
the Pietersburg could have been regarded as one of the better
known industries of the MSA (Sampson 1972:p.63),many
questions remain today regarding its idiosyncrasy and chrono-
stratigraphical context. Such questions include notably the
following: does the Pietersburg truly correspond to a distinct
technological entity, which are its main techno-typological
characteristics, and how long did it last?
Second, as summarised by Wadley (2015), it seems that the
oldest MSA sites in southern Africa are located in the interior
part of the region. Investigations about the Pietersburg could
therefore contribute to shed some light on the beginning of the
MSA in southern Africa.
Third, it is certainly pertinent to attempt to propose a dif-
ferent picture from a somewhat simplistic view according to
which the Still Bay and Howiesons Poort techno-complexes
would be, respectively, preceded and followed by the monot-
onous pre-Still Bay and post-Howiesons Poort phases (Will
et al. 2014). Interestingly, neither at Mwulus Cave nor at
Cave of Hearths or Bushman Rock Shelter (two other sites
located in Limpopo Province, South Africa) have these
techno-complexes been identified.
Fourth, inquiry into MSA variability in this area allows us
to link it with other African regions where the MSA is still
poorly described, such as in Zimbabwe, Botswana and
Mozambique. Notably, most of the hypotheses put forward
in the last decade on the MSA have originated mainly from
coastal sites, such as the cape floral hypothesis (Marean 2010,
A final issue, recently mentioned by Porraz et al. (2018), is
that if the Pietersburg corresponds indeed to a local variation,
then true techno-cultural regionalization would exist since the
early stages of the MSA, as proposed by Clark (1959).
Mwulus cave: discovery of the site and Tobiass
excavation (1947)
Serendipity: this ishow P.V. Tobias described in his memoirs
his discovery of MwulusCave(Tobias2005). Brian Maguire
had decided to show a group of students from the University
of the Witwatersrand a yellowwood tree at the entrance of a
cave in a remote location on his family farm in Limpopo.
Amongst this group of students was a young P.V. Tobias,
who, just after they had climbed to the site, found some stone
tools scattered on the surface of this small cave. Soon after, in
1947, Tobias decided to excavate the cave they had found with
B. Maguire, a rather disappointing excavation for the trainee
palaeoanthropologist, given the lack of bone preservation.
Thenameofthesite(MwulusCave) was given by
Constantine Duncan Maguire, the former owner of the farm
where the cave stands, in memory of a hermit called Mwulu,
who lived in the cave in the nineteenth century (pers. comm.
Judy Maguire, 2016). Indeed, during his excavations, Tobias
found some large pottery sherds that he assumed belonged to
Mwulu, the hermit.
Tobias did not leave any information about the precise
location of the site, other than that is was located near the
famous Makapan Valley in the Makapansberg Mountains. In
van Riet Lowes archives housed at the Origins Centre,
University of the Witwatersrand, there were some coordinates,
which correspond to the boundary between the Spanje 36KS
and Portugal 55 KS farms.
Mwulus Cave is located in the eastern escarpment of the
Makapansberg Mountains in the district of Mokopane
Archaeol Anthropol Sci
(Limpopo Province), very near the Makapan Valley (around
5 km away from Cave of Hearths; Fig. 1). The cave is situated
directly in a mountain pass, informally called the Monkey
Nekby the Maguire family. It is north facing and accessing
it requires climbing an ca. 15 m high quartzite cliff (Fig. 2).
Tobias excavated half of the cave deposits. Even though he
did not specify which half it was in his 1949 paper, it is likely
that it was on the left-hand side as one enters the cavity
(Fig. 3). This is quite evident nowadays as this is where the
bedrock is visible. For his excavation, he established a grid,
following Malans excavation method. Unfortunately, the ex-
act location of his squares is unknown as he did not provide
any spatial information, in either the archaeological material
(currently stored in the basement of the Origins Centre at the
University of the Witwatersrand, Johannesburg), or in his two
papers on MwulusCave(Tobias1949,1954). The only in-
formation available to us today is the name of the squares, not
their location (see Table S1 and Supplementary material).
In his first paper on the site, Tobias described the surface
layer (surface rubble) overlying five more layers, namely
(from top to bottom): Ash and Sand III,Red Sand,Ash
and Sand II,Red Sandand Ash and Sand I(Fig. 4a).
According to his description, the red sand layers were virtually
sterile, whereas the ashy and sandy layers were rich in archae-
ological lithic material. Nonetheless, bone remains are not pre-
served at all, most likely because the cave formed in quartzite.
Besides the few pottery sherds found in the surface layer and
attributed to the hermits occupation of the cave, the entirety of
the archaeological assemblage is composed of lithic artefacts.
Even if Tobias had adequately described the stratigraphy
illustrated in his 1949 paper (Fig. 4a), the labelling of the
archaeological material at the Origins Centre reveals that he
excavated by spits, measured in inches and followed irregular
divisions. Moreover, he grouped these spits in three Beds (I,
II and III), I being the lowest. A synthesis of the relationships
between the different spits and various stratigraphic beds is
provided in the Supplementary Material and in Table S1.
Those divisions highlight two issues concerning the relation-
ship between the stratigraphy on the one hand and Tobiass
excavation methods on the other. Firstly, the fact that all spits
are strictly continuous means that it is more than likely that
some of them are lumping the three ashy and sandy beds with
the inter-stratified two red sands layers. In other words, if the
red sands were actually sterile, as Tobias pointed out in his
1949 paper, one would expect that some spits would contain
no material. This is not the case, thus suggesting that some
mixing occurred, because of the excavation methods rather
than stratigraphic processes. Secondly, his labelling of the
lithic material shows that Beds III and II share some spits,
notably spits 12-24 and 12-25. With no further notes or expla-
nations, we cannot know if this corresponds to the bottom part
of Ash and Sand III,theUpper Red Sandor the top part of
Ash and Sand II.
Regarding Tobiass excavation methodology, it is clear that
he did not collect the chips and it is likely that he selected the
material stored at the Origins Centre. As an argument to sup-
port this statement, our two campaigns yielded 1700 lithic
pieces over 2 cm and around 15,000 chips in an area of
170 × 40 cm (0.68 m
), whereas Tobiass excavation targeted
a much larger area (around 25 square yards or 20.9 m
cave but yielded only3000 pieces. Furthermore, the prelim-
inary analysis of the old collection has revealed that quartz
blanks seem particularly underrepresented in his collection.
Mwulus cave: geological, geomorphological
and ecological setting of the site
Geological setting
Mwulus Cave is hosted within the middle beds of the Late
Archaean, Early Proterozoic (26422584 Ma) quartzites of the
Black Reef Formation (BRF), which are considered the basal
lithostratigraphic unit of the Transvaal Supergroup exposed
around the margins of the Transvaal basin (Eriksson et al.
1995;Elsetal.1995). In the area of Mwulus Cave, the BRF
unconformably overlies Archaean basement rocks and the
Godwan Formation and conformably overlies the Wolkberg
Formation (Button 1973;Erikssonetal.1995). The BRF is con-
formably overlain by, and in some areas grades into, the dolo-
mites of the Malmani Subgroup (Button 1973,1986;Eriksson
et al. 1993,1995). Locally, the BRF quartzites are described as a
sequence of trough and planar cross-bedded quartzites and sand-
stones with interbedded conglomeritic units (Button 1973;Henry
et al. 1990;Erikssonetal.1993,1995). Quartzites are generally
medium to coarse-grained and light coloured (Maguire 2009).
Genetically, the BRF sediments formed under fluvial conditions,
with the lower, upwards-fining sequences, interpreted as either
subtidal, marginal marine-fluvial sand sheet settings (Button
1973; Key 1983,1986; Eriksson et al. 2001). The upper,
coarsening-upwards sedimentary sequences are interpreted as
forming in a braided delta, possibly lacustrine environment
(Henry et al. 1990;Erikssonetal.2001).
The south-west tilted, heavily fractured BRF forms the NNW-
SSE orientated Highlands Mountains of the Makapansberg
eastern escarpment overlooking pre-Wolkberg Archaean for-
mations of the Springbok Flats to the east (1524 average
masl.). The overlying dolomites have been weathered back
from the escarpment edge, resulting in extensive exposures
of resistant quartzite as the highest topographical feature (up
to 2040 msl) (Maguire 2009). The western slopes of the es-
carpment are heavily faulted but less abrupt, tilting with the
bedding of the BRF and overlying dolomites and carrying
water away from the escarpment (Maguire 2009). The scarp
Archaeol Anthropol Sci
Fig. 1 aLocation of Mwulus Cave (Limpopo, South Africa). bDigital
model of terrain with Geographical coordinates and Makapan Valley,
Mwulus Cave and Mokopane indicated. The digital model of the
terrain was retrieved from maintained by the
NASA EOSDIS Land Processes Distributed Active Archive Center (LP
Archaeol Anthropol Sci
near the opening to Mwulus Cave presents a range of verti-
cally variable quartzitic facies including medium to coarse-
grained sediments with cross-bedded, upwards-fining and
massive structures. The cave formed in a coarse-grained (
1.0 to 0.5 Phi) massive quartzite, which is exposed in the
walls, roof and floor.
In the area of Mwulus Cave, significant NE to SW frac-
turing has formed steep fault-controlled valleys and cliffs
dissecting the scarp (Maguire 2009). It is along a similarly
orientated vertical fracture system that the cave formed in
the exposed cliff face.
Mwulus Cave is located 30 m below the overhanging
quartzite ledge, ca. 15 m above the base of the cliff and 5 m
below a compound fracture that branches into two diverging
joints above the roof of the cave. It is most likely that the cave
formed through the localised collapse of the quartzite as a
result of the development of this fracture complex. Two small-
er cavities have formed as a result of the same process several
meters to the left and right of Mwulus Cave. The lateral extent
of Mwulus Cave is constrained by the limits of the fracture
complex and is about 8 m wide at its opening. Abundant
vertical joints in the caves roof and walls facilitate cave break-
down and fluid penetration all the way to the back of the
visible cave. Longitudinally, the cave extends into the cliff at
least 14 m in a SW direction and the roof and walls taper
towards the back of the cave (Fig. 3)toawidthof4mand
height of 50 cm respectively above the deposit surface.
Although not explored past this point, the cave appears to
extend further several meters and is almost completely filled
with collapsed blocks and sediments. Exposures of the floor of
the cave are limited to excavations by Tobiass(1949) and this
project, both of which suggest that the floor is irregular in
morphology with occasional boulder-sized blocks and abun-
dant medium (10 cm) tablets and angular chunks of dipping
quartzite detached from the floor but decaying in situ.
Abundant small to medium (510 cm) tablets and angular
chunks are found on surface of the deposit especially near
the northern wall. Small tablets are seen exfoliating from the
walls and roof of the cave. As Tobias (1954) notes, the walls
and roof (especially in the northern part of the cave) are cov-
ered in a thin coating of speleothem, which has also locally
calcified sediments and formed thin, punctuated flowstone on
surficial sediments and on the undersides of clasts to a depth
of 10 cm (Fig. 4). Tobias (1954) attributes this speleothem
growth to the last phase in the depositional sequence of
Although there is no evidence of flowing water within the
cave, moss, grasses and small woody plants grow all the way to
the back and sediments are generally moist to the base of the
deposits. Tobias notes that during his visit abundant water was
dripping from the walls in the recesses of the cave(1949:3).
Outside the cave, a platform extending about 4 m has
formed on very large collapsed quartzite blocks. Shallow soils
trapped by these blocks create a generally flat platform sur-
face, which is well vegetated. This has probably been signif-
icantly supplemented by Tobiass sieve pile. The platform
ends abruptly in a 15-m cliff above a steep but stable scree-
slope composed of quartzite boulders. Access to the cave is
achieved by climbing up the edge of this platform and is likely
to have limited entry to animals with climbing proclivities.
Abundant small mammal spoor are present in the cave, as
are their bones, suggesting regular small mammal occupation
(especially the rock hyrax Procavia capensis) and visitation
by small carnivores and owls.
Fig. 2 aMountain pass, informally called the Monkey Nek, which
leads to Mwulus Cave. bMwulus Cave in the Makapansberg
Mountain range. cInterior of the cave in September 2017
Archaeol Anthropol Sci
Fig. 3 Topography of MwulusCave.aZenithal view. b3D with excavation areas. In purple: illegal test pit found in 2016. In green: area excavated in
2017. Topography: F. Colino
Archaeol Anthropol Sci
Climate and vegetation background
Located in the northern Highveld of South Africa, the
Limpopo Province has been characterised by summer rainfall
conditions throughout the Pleistocene, unaffected by the per-
sistent fluctuations in the position of the winter rainfall zone in
the southern half of the country during the Quaternary (cf.
Thackeray and Fitchett 2016; Fitchett et al. 2017). The sum-
mer climate is dominated by convective storms in the late
afternoons, while the dominant high pressure over the interior
results in atmospheric stability throughout the winter months
(Tyson and Preston-Whyte 2005). The study region receives a
mean annual precipitation of ~ 520 mm/year, of which a sub-
stantial 90% falls between the months of October and March
(Repinski et al. 1999). Mean annual temperature for the region
is in the range of 1819 °C, with a range from mean minimum
temperatures of 2.4 °C in winter to mean maximum
temperatures of 27 °C in summer (Repinski et al. 1999;
Stevenson et al. 1999). Due to the clear skies and high diurnal
temperature variation, frost occurs occasionally during the
winter months (Rayner et al. 1993).
Mwulus Cave is located in the Savanna biome of South
Africa. The vegetation of the region has been mapped at
1:250,000 scale by Mucina and Rutherford (2006), and we
use this assessment for identifying and describing vegetation
units within our study area. Two savanna vegetation units
dominate in the area, the Polokwane Plateau Bushveld in the
high-lying plains west of the site and Mamabolo Mountain
Bushveld on the hills and small mountains of the area (after
Rutherford et al. 2006). Patches of the Strydpoort Summit
Sourveld vegetation unit belonging to the Mesic Highveld
Grassland bioregion of the Grassland biome are also present
along rocky summits and mountain slopes (Rutherford et al.
2006). The hills and low mountains are very rocky, covered by
Fig. 4 aTob iass stratigraphy
(Tobias, 1949). b.MwulusCave
West profile stratigraphic se-
quence with total station-derived
depths on the right vertical axis.
The left vertical axis presents
Tob iass stratigraphic divisions
and nomenclature (1949, 1954)
and the Units identified in the
field and excavated in 2017 (IV)
Archaeol Anthropol Sci
small trees and shrubs with several succulents. On the high-
lying plains, the vegetation is characterised by a short open
tree layer with a well-developed grass layer with occasional
trees at higher altitudes (Rutherford et al. 2006). C
dominate the graminoid component, mainly belonging to
three Poaceae subfamilies, which are, in decreasing abun-
dance, the Panicoideae, Chloridoideae and Aristidoideae.
New excavation: results and analyses
New excavation protocol, sampling and analytical
When we accessed the cave for the first time in September
2016, there was a test pit partially covered by a deteriorated
large plastic sheet, which did not seem to correspond to
Tob iass excavation. The size of this test pit was approximate-
ly 1.70 cm long by 85 cm wide and 1.5 m deep (Fig. 3b, see
square in purple). In the first excavation campaign (in
June 2017), the first task was to clean this putative illegal test
pit to access the exposed profiles and compare them with
Tob iass stratigraphic description. Once the four profiles were
cleaned, it became clear that this half of this putative illegal
excavation only damaged the backfill of Tobiass excavations.
Although, it must have cut unexcavated deposit at least ~
15 cm to the west. On the west profile, the stratigraphy seemed
to correspond to Tobiass description. We decided to excavate
a 40-cm area into this profile in two squares: W1 and W2
(Figs. 3and 4). We established these two squares in order to
have a general location for all the chips and small archaeolog-
ical remains measuring under 2 cm. These two squares occupy
a total surface of 1.70 m × 40 cm. Tobiass excavations ex-
posed deposits to a depth of 144 cm (57 in) at Mwulus Cave.
Our excavations reached a comparable depth of 127 cm from
the current deposit surface to the cave floor. During the second
campaign, we continued these excavations until we reached
the floor of the shelter. We plotted all archaeological finds
measuring over 2 cm with a Nikon Nivo 5C Total Station.
As some of the layers lacked discernible internal stratigraphy,
we subdivided them into successive plans during excavation,
following the natural stratigraphy. All sediments were dry-
sieved and labelled by square.
Fabric information was recorded for all natural or
artefactual pieces larger than 5 cm with an elongation ratio
of at least 1.7:1 (L/W) (following Bertran and Texier 1995)
and with a flat underside surface. Orientation was divided into
cardinal points (N, NE, E, SE and S), and dip was measured
with a Brunton geological compass recording the direction of
dip using the same cardinal points. The data yielded three
possible fabric variables, orientation, dip and dip direction
(if different from orientation).
Several photogrammetry models were produced using
Agisoft software, notably one recording the whole site and
another one documenting the new profile unveiled by the
new excavation.
The topography of the cave surface and ceiling of the cave
were documented with Surfer software (v. 12.0.626) (Fig. 3).
In both processes, the interpolation of points was accom-
plished using the Kriging method.
We took 29 bulk samples for sedimentology, pollen and
phytolith analyses (specific methodology is presented in the
Supplementary material).
Four micromorphology blocks were extracted at the con-
tacts between Units I and II, II and III and III and IV. Analysis
of the blocks is ongoing.
Established methods and criteria for the physical analysis
of archaeological ochre were used to categorise the assem-
blage (e.g. Watts 1998; Hodgskiss 2012). Pieces were classi-
fied by geological form, grain size, hardness, colour and spec-
ularity (mica-inclusions), and the presence of use-traces and
post-depositional markings was recorded.
For the chronology, 12 samples were taken for optically
stimulated luminescence analysis. These samples were sepa-
rated into two sets of six samples each, and sent to two differ-
ent laboratories. The preliminary results were completely con-
tradictory and alternative methodologies are being applied in
order to understand these discrepancies and in order to obtain
an accurate optically stimulated luminescence chronology of
the site. These results are not presented here as these analyses
are ongoing. In addition, three burnt lithic samples were col-
lected during excavation for thermo-luminescence dating.
The lithic technology analysis followed the methodology
elaborated upon by de la Peña (2015), which follows the
chaîne opératoire approach combined with simple controlling
statistical tests. The spatial analyses of the lithic distribution
were performed with QGIS (v. 2.18.11), ARCGIS (v. 10.1)
and PAST.
The present day surface topography of MwulusCave(Fig.3),
inside the dripline, may not accurately reflect its original state.
Tob iass excavation was orientated longitudinally, extended in
one yard squares (presumably in a trench configuration) and
confined to one half of the cave floor, leaving the other intact
to provide a witness section(Tobias 1949: p. 4). Through the
excavation process, some areas may have been trampled flat,
sediments may have been displaced to the sides of the cave,
and the trench itself may have suffered post-backfill subsi-
dence. When this project started, laterally, the deposit has a
shallow concave profile, with a central depression orientated
Archaeol Anthropol Sci
longitudinally and extending from just inside the drip line to
the back of the cave. Close to the walls, and out of reach of
animals, the surface was flat. Surficial water runoff from the
drip line into the central area of the cave was evident.
Stratigraphy and sediments
The excavation of Mwulus Cave identified five stratigraphic
units, named I to V from the surface to cave floor (Fig. 4b).
These units were defined in the field prior to and during
excavation through close inspection of vertical and lateral
sedimentary characteristics including colour, content, texture,
structure and clast fabric. Within defined units, very little to
no internal stratigraphy was observed. It must be noted that
units I to V are not identified as distinct facies and the nature
of the boundaries identified in the field and the stratigraphic
division of the sequence is discussed later. Units I to V
compare closely to those identified by Tobias (1949;1954;
Fig. 4a, b) although unit boundaries were more diffuse than
illustrated in Tobias (1954). Sedimentological analyses were
conducted to test the accuracy of the identified units with re-
spect to Tobiass(1954) hypothesis that his layers A to Eand
strata1to5(our I to V) represented distinct phases of fluctu-
ating aeolian sedimentation in the shelter.
Throughout the Mwulus sequence, sediments are sandy,
unconsolidated and without structure. Textures vary laterally
and vertically locally but can be considered sandy to loamy
sand (granulometry is discussed below). Sediment colour
varies from pale grey to very dark brown and reflects varying
abundances of organic matter and moisture. Inorganic carbon
varies between 0.4 and 1.2% content with the highest content
found in Unit III. Only one clearly conformable structurally
defined contact was found which separated Unit I from II;
otherwise, deposit contacts are diffuse and structural intra-
unit stratification is not discernible. Infrequent laterally punc-
tuated horizontal dark lenses with diffuse upper and lower
contacts do appear in the deposit and represent concentrations
of burnt plant matter.
Sedimentologically, no significant distinction can be made
between the field-identified units (and by association Tobiass
strata), and analyses of bulk samples presented here provides
useful insights into the nature of accumulation of the Mwulus
sediments. The matrix of the Mwulus deposits ranges from
coarse sands to medium silts with a dominance of sand-sized
particles and relatively minor silt faction that contributes tothe
loamclassification of the sediment texture. Figure 5shows
particle size distribution curves for sediment samples
representing the different stratigraphic units identified through
the Mwulus sequence from deposit surface to cave floor
(units I to Vrespectively). For comparative purposes, a sample
from the decaying wall of the cave is presented, as is the
identified particle size range for the Kalahari sands (Schlegel
et al. 1989). Through the Mwulus sequence a consistent and
proportionally similar contribution of two granulometrically
distinct sediments (represented by the bimodality of the parti-
cle size distribution curves in Fig. 5)isdemonstratedinthe
sand component (Fig. 6).
Autogenic contributions to the matrix are simple to spot
within the bimodal particle size distribution curves, contribut-
ing particles of between 1and1.0Φ, and are generally
consistent with a very coarse faction contributing up to 5%
of the sands (Fig. 6), and coarse faction contributing 15 and
23%, suggesting an even rate of cave breakdown over the
depositional sequence. The medium and fine (2 to 3.5 Φ)s
component represented in the finer distribution peak matches
closely with the size range expected of an aeolian provenance.
The most likely source of aeolian grains being the Kalahari
sands (proposed as a significant contributor to the cavessed-
iment by Tobias (1954) (see comparative particle size range in
Fig. 10; from Schlegel et al. 1989), a significant aeolian con-
tributor to the southern African landscape after 2.4 million
years ago (Partridge 1993). Given the dramatic elevation of
Mwulus Cave above the southern and eastern Springbok
Flats (1524 average msl), it is to be expected that the generally
finer (3.5 Φ) faction of the Kalahari sands were deposited in
the cave. Ongoing microscopic analyses of the sediments,
Fig. 5 Particle size distribution
curves in Phi Units (Φ)of
representative samples for each
identified stratigraphic unit in
Mwulus Cave. Data are distilled
from quantitative granulometry
yielded from a Mastersizer 3000
Archaeol Anthropol Sci
including micromorphology and SEM, will help confirm the
provenance of the finer sand particles. The allogenic sands (2
to 3.5 Φ) consistently contribute 3540% of the total sands but
do not vary through the sequence to a degree that would sug-
gest distinct episodic fluctuations in aeolian deposition
(contrary to Tobias 1954).
Natural clasts derive from the autogenic decay of the cave,
and generally, clast abundance is low through the sequence,
except at the base where abundant unsorted tabular natural
clasts are found in varying states of decay associated with
the shelter floor. Minor additional local lithologies are repre-
sented as clasts in the deposits in the form of shales, mudstone,
siltstones, sandstones, quartz crystals and iron oxide nodules
in various forms. Once deposited, the tabular coarse-grained
quartzite clasts decay in situ, rounding rapidly and contribut-
ing angular coarse sand particles to the matrix. In numerous
instances, discrete, unconsolidated light grey patches of coarse
sand were identified during excavation and represent
completely decayed autogenic clasts that have remained
Biotic and chemical contributions to the Mwulus Cave
sediment suite include abundant burnt and unburnt plant ma-
terial, represented as very small to medium (< 1 cm diameter)
roots distributed in fluctuating quantities throughout the de-
posits. The distributions of these components can be seen in
the distribution of moisture and organic carbon content
through the sequence (Fig. 7). There is a close association
between areas of high artefact abundance and moisture and
organic carbon content. All three lower units (III to V) show
higher organic carbon contents than units I and II. For exam-
ple, Unit III is dark brown, wet to the touch and comparatively
rich in organic carbon, with roots often directly associated
with artefacts and clasts. Unit IV, although demonstrating a
drop in moisture from unit III, shows a relatively high organic
carbon content, and bothmoisture and organic carbon increase
significantly in to unit V, where natural clast abundance is
high. The consistency of poorly consolidated sandy sediments
through the sequence suggests water (and by association plant
growth) has accumulated in areas of increased artefact and
clast abundance.
Larger biotic contributions include modern bones and fur
of visiting animals and their prey, althoughthe rare occurrence
of bones is limited to the uppermost unitthe low pH of the
host quartzitic sediments destroyed older faunal material.
Animal faecal matter and dust from visiting animals, in addi-
tion to ash and decomposed charcoal, probably contribute to
the majority of the silt faction (from 4 to 7 Φ). Biogenic struc-
tures are present in the eastern wall and have significantly
disturbed the deposits. These biogenic structures have various
sizes and, therefore, could be attributed to different agents
(Backwell et al. 2012).
Post-depositional modifications will be studied in detail
through the ongoing micromorphological analysis, but here
there is an opportunity to propose hypotheses based on the
Fig. 6 Variation in proportional
contribution of different sand to
coarse silt factions (in Φand μm)
through the MwulusCave
sediment sequence.
Granulometric data yielded from
Mastersizer 3000 analyses of all
sediment samples. Blue data
represents the autogenic
contribution of very coarse and
coarse sands and red data
represents the contribution of
medium to fine sands
Archaeol Anthropol Sci
sedimentology and organisation of the artefacts and clasts ex-
cavated from the units. In the central part of the chamber,
numerous borrows, anthropogenic cuts and fill structures are
seen, which is why excavations focused on the more intact
western wall where no such features could be seen. Figure 8
presents a rose diagram and stereonet projection of orienta-
tions and dips of elongated and flat-bottomed clasts and arte-
facts excavated (n= 131). Although a large enough sample
was not present for a statistical comparison between units,
the data, and observations during excavation, indicate gener-
ally flat depositional surfaces and an absence of directional
processes (isotropic orientations). Vertical biogenic processes
(i.e. trampling, burrowing, root disturbance indicated by
strongly dipping pieces) affecting the larger components of
the deposits is also minimal. It can be suggested that although
bioturbation may have affected finer factions of the sediments
and probably blurred unit boundaries and destroyed ancient
anthropogenic features, artefacts and clasts probably remain in
close proximity to their original organisation and distribution
in the deposits.
Pollen analyses
A very low pollen concentration was observed for the majority
of samples (Table 1). For two samples, IIIa and IIe, higher
pollen concentrations were recorded, but significantly lower
than the 300 grain threshold required for statistical analyses.
No pollen grains were observed in the analysis of the slides
from samples Id, Ie, If, IIIc, IVc and Va. The variability in the
number of pollen grains preserved per sample could be ex-
plained by differences in production, dispersal, deposition or
preservation; each of these factors is significant in potentially
representing stratigraphic distinctions and is likely driven by
variations in regional climate. It is difficult, however, to deter-
mine the proximal cause for the variation in abundance or to
assign a causal factor, without a greater and more representa-
tive pollen distribution.
For the samples containing fossil pollen, the most common
pollen that was identified throughout each of the layers includ-
ed cosmopolitan families, notably Cyperaceae, Poaceae and
Asteraceae (Table 1). Two samples contained smaller
Fig. 7 Histogram of proportional
moisture, organic carbon and
carbonate content vertically
through the MwulusCave
sediment sequence from loss of
ignition (LoI) analyses. Data rep-
resented includes only vertically
associated samples and excludes
laterally associated samples
Fig. 8 Rose diagram and
stereonet projection of elongated
clasts (n=131)excavated
through the MwulusCave
sediment sequence during the
2017 campaign. The rose diagram
was generated using
freeware and the stereonet
projection was generated using
Stereonet 10.1 freeware based on
Allmendinger et al. (2013)
Archaeol Anthropol Sci
Table 1 Raw pollen counts from MwulusCave
Sample no. Asteraceae Poaceae Cyperacea Combretac Proteaceae Malvaceae Apiaceae Liliaceae Cheno-
Crassula Pinus Grewia Spores Microchacoal Algal material
Ia 1 4 YY Y
Ib 1 YY Y
Id No pollen YY Y
Ie No pollen YY Y
If No pollen YY Y
IIa 1 YY Y
IIb 1 1 YY Y
IIc 1 1 YY Y
IId 1 1 YY Y
IIe 6 68 33 5 4 1 10 Y Y Y
IIf 9 16 12 2 8 1 2 Y Y Y
IIIa 7 7 28 1 1 1 1 2 Y Y Y
IIIb 2 1 1 Y Y Y
IIIc No pollen YY Y
IIIe 1 1 YY Y
IVa 10 4 1 1 1 Y Y Y
IVc No pollen YY Y
IVd 1 1 1YYY
IVe 1 YY Y
IVf 14 2 1YYY
Va No pollen YY Y
Vb 1 1 YY Y
Vc 1 YY Y
Vd 1 YY Y
Archaeol Anthropol Sci
proportions of fossil pollen from the Combretaceae,
Malvaceae, Liliaceae and Cheno-Am group families and of
Crassula,Grewia and Pinus species (Table 1). Spores and
algal materials were found on slides from all of the samples,
including those for which no pollen grains were found, indi-
cating high humidity suggestive of the persistent infiltration of
water into the cave throughout the time period represented by
the sample material. Microcharcoal grains were also found in
all samples. Due to the light weight of these particles, they
could be sourced from a large catchment area. Moreover, the
timing of the deposition of these charcoal grains cannot be
ascertained from the pollen assemblage.
Notably, each of the higher-concentration samples
contained Pinus pollen grains, indicative of contamination
from the contemporary environment, as this species was in-
troduced to South Africa ~ 300 years ago. Pinuspollenwas
also identified in samples Ve, IVf and IIb, indicating contam-
ination from modern pollen in samples throughout the se-
quence. The variation in the total number of preserved pollen
grains, and the differential degrees of degradation of individ-
ual grains, would however prohibit a conclusion that all pollen
grains from the excavated material are modern contaminants.
Here, we present the preliminary results of the phytolith study
from the entire deposit at Mwulus Cave. The ongoing study
will include the study of modern surface soil samples from the
cave surface and from the vegetation surrounding the cave. A
more in-depth taphonomic analysis of the assemblage will
also be provided in the future.
There is an increase in phytolith concentration from the
lowermost layers (V to III) to the uppermost ones (II and I)
(Fig. 9). Sample IIe showed the highest phytolith concentra-
tion amongst samples (4.2 million phytoliths /g of sediment).
Together with phytoliths, diatoms and sponge spicules were
also identified in most of the samples analysed, being an in-
dication of moist soil conditions in the cave. The abundance of
these bio-siliceous microfossils increases accordingly with
that of phytoliths, and it could be indicative of differential
preservation conditions at the site.
A total of 64 phytolith morphotypes were identified in the
29 archaeological sediment samples. These were later grouped
by plant types and plant parts into ten general categories:
grassesPoaceae, sedges-Cyperaceae, leaves and wood/
bark of dicotyledonous plants, spheroids, epidermal append-
ages, elongates with and without decorated sides, blocky mor-
phologies and irregular and indeterminate morphologies.
The phytolith assemblage from Unit V was characterised
by a low abundance of grass phytoliths, and the dominance of
spheroid phytoliths, blocky and thin parallelepipeds, and ir-
regular morphologies which have all been traditionally asso-
ciated with wood/bark (Albert 2000; Albert and Weiner 2001;
Collura and Neumann 2017; Esteban et al. 2017a,b;Murungi
2017; Tsartsidou et al. 2007). These morphotypes are amongst
the most resistant phytoliths and tend to dominate phytolith
assemblages affected by post-depositional processes (Cabanes
et al. 2011; Cabanes and Shahack-Gross 2015); therefore, the
phytolith results from unit V will not be interpreted until more
in-depth taphonomic studies are conducted.
Conversely, the phytolith assemblage from the uppermost
levels (Fig. 10) was dominated by the presence of grass
phytoliths, mainly grass silica short cells (GSSCs) from the
lobate and rondel types. The grass phytolith assemblage
showed considerable differences amongst samples from units
I and II. The former showed the highest presence of GSSC
rondels, typically associated with C
grasses from the
Pooideae, Ehrhartoideae and Danthonioideae subfamilies,
and the lowest of lobates, which are generally common
Panicoideae in South African records (e.g., Cordova 2013;
Cordova and Scott 2010; Esteban et al. 2017a,b;Murungi
2017;Rossouw2009). Conversely, C
graminoids (C
and sedges) dominated the phytolith assemblage in unit II.
Samples from unit II was characterised by the domi-
nance of GSSCs lobates, followed by high frequencies of
rondels and saddles, sedge phytoliths (papillae/hat shape
Fig. 9 Phytolith concentration (phytoliths per gram of sediment) at the different levels studies from Mwulus
Archaeol Anthropol Sci
and achene morphotypes), blocky parallelepipeds and
elongates without decorated sides are also well represent-
ed in the phytolith assemblages. Conversely, the phytolith
assemblage of unit I was characterised by the dominance
of elongates without decorated sides, GSSC rondels and
blocky morphologies (polyhedral and parallelepiped),
followed in abundance by GSSC lobates, prickles, spher-
oids and irregular morphologies. Samples from this level
showed the lowest frequencies of sedge phytoliths (papil-
lae/hat shaped and achene morphotypes).
The 2017 Mwulus Cave excavations uncovered a total of
356.3 g of ochre. Of that, 189 g is attributed to pieces <
10 mm in maximum length. All pieces 10 mm were
analysed, totalling 137 pieces with a weight of 167.3 g.
Five geological forms constitute the assemblageshale,
ferruginous mudstone, ferruginous siltstone, ferruginous
sandstones and iron oxide. The analysed pieces comprise
mostly shales (56.2%), followed by mudstones (21.2%) and
siltstones (11.7%), with only four iron oxide (or haematite)
pieces in the assemblage. Just over half of the assemblage
(56.2%) is soft (Mohs 2 and below), and 40.1% have medi-
umhardnessvalues (Mohs 3 and 4). Most of the pieceshave a
small grain size, with 64.2% of the pieces with either clayey or
mixed clayey-silty grain sizes. Few of the pieces (21.2%) are
mica-rich which gives them a specular, sparkly quality. Colour
was grouped into seven categories (see Hodgskiss 2012)
brown and grey. Slightly less than half the assemblage
(44.5%) has red hues, with unusually highquantities of yellow
hues and oranges (54.7%) compared to other MSA sites.
The percentage of utilised pieces is lower than average for
MSA sites (roughly 1015%) with only three pieces showing
signs of utilisation (2.2% of the assemblage) (Fig. 11ac). All
the utilised pieces are bright red, which is consistent with the
general MSA preference to use bright red ochre (Henshilwood
et al. 2001,2014; Watts 2002; Hodgskiss 2012; Dayet et al.
2013). Two are specular iron oxide pieces and one is a mud-
stone. All have medium-hard hardness values (Mohs 4 and 5)
and are fine grained. They appear to have been rubbed against
soft materials (like skin or hide) resulting in the formation of
microstriations, smoothing and polish. No striations that
would have formed if they were ground against a hard rock
surface are apparent.
The bulk of the assemblage derives from units I and II
(30.7% and 42.3% of the assemblage respectively). The three
utilised pieces are from units I, II and IV, with the utilised iron
oxide pieces appearing in the younger layers (I and II). The
piece appearing in the older unit (unit IV, Fig. 11b) is the only
piece from the 2017 excavations that has clear grinding stria-
tions. No mica-rich, specular pieces (utilised or unutilised) are
found in the older layers (units IVand V). There is no signif-
icant variation in the types of ochre collected through time.
Previous ochre studies
Amongst his finds, Tobias focused particularly on different
fragments of specularite and proposed that the use of this
material in the MSA people would have been for artistic pur-
poses: Well-rubbed pieces of specularite suggest the artistic
practices of the Pietersburg folk(Tobias 1949: p. 10). Watts
(1998) analysed the ochre excavated by Tobias. Like him, we
found 13 pieces of ochre (termed pigmentby Watts) with a
total weight of just under 500 g (our weight = 497.1 g,
Watts= 477.6 g). These 13 pieces together weigh almost
Fig. 10 Histogram showing the phytolith morphological distribution, by plant types and plant parts and Grass Silica Short Cell, in samples from
Mwulus Cave amongst the different stratigraphic units
Archaeol Anthropol Sci
three times more than the entire 2017 ochre assemblage.
These pieces are mostly heavy iron oxideshaematite and
speculariteand they are generally larger pieces than the
2017 assemblage, with 12 of the 13 pieces over 30 mm in
length. It is likely that only the bigger and more impressive
pieces would have been kept, with a large portion of the small
pieces being thrown away by excavators or unidentified. We
agree with Wattss analysis of the utilised assemblage apart
from one piece. We identified six pieces with use-traces while
Watts identified seven; the differing one piece we could not
confidently identify as utilised. Four of the utilised ochre
pieces are from the 1224 in spit, and two have unknown
stratigraphic provenance. The utilised pieces are mostly
ground, some with clearly faceted surfaces and crayon-
shaped tips (Fig. 11df). Some of these ground pieces have
polish and microstriations on the ground surfaces, possible
evidence that they were rubbed on a soft material after they
were ground.
Spatial analyses of the lithic artefact distribution
The 2017 excavations sampled the full stratigraphy from the
surface until bedrock. As can be seen in the plotting of the
lithic artefacts (Fig. 12a), the distribution-density of the lithic
artefacts does not correspond with the five units described
during our excavation. Indeed, as illustrated in Fig. 12a, it
seems that instead of the five clusters (that would be expected
following the five stratigraphical units defined in the field),
there are only three clusters regarding the lithics: a first one
consistent with unit I, a second one with unit II and a third one
with units III-IV-V lumped together (Fig. 12b, c). The lithic
distribution of lithics every 10 cm in depth (Fig. 12d) also
indicates three distinct clusters.
In order to test this further, we performed different spatial
analyses to see if the lithic material from the five stratigraph-
ical units cluster or not spatially. Firstly, we used the Morans
I, which is a measure of spatial autocorrelation. This analysis
determines whether the distribution is positively correlated
(that is, there are preferential clusters, values close to 1), neg-
atively (without any association since they are uniformly dis-
tributed, values close to 1) or is random (values close to 0).
We performed this test for all the lithics, plotted in the five
units together. The first task performed was building up a grid
f(Table2) for determining the randomness of events; in this
case, the lithic distribution. From this grid, we ran a Spatial
Autocorrelation Analysis, and the results clearly show that
there is a cluster distribution being the Z-score 12.18 and the
I Moran value 0.64 (positive autocorrelation) (Table 3).
Clustering was also demonstrated following the average
nearest neighbour analysis, being the Z-score 15.93 call to
Table 4; thus, it is highly unlikely that the distribution re-
sponds to a random pattern. Finally, a hotspot analysis was
carried out; for this, the Kernel density estimate was used
(Getis and Ord 1992)(Fig.12c). This analysis is the
Fig. 11 A selection of utilised ochre pieces from MwulusCave
(2017 and Tobiass excavations). aA hard, specular iron oxide
nodule (MWO_049 from Layer 1, square W1, 2017 excavations)
that has smoothing, rounded edges and microstriations, indicating
it was rubbed against a soft material. bA hard, iron oxide nodule
(MWO_095 from Layer II, square W2, 2017 excavations)
displaying microstriations and smoothing on the surface, indicating
that the piece was rubbed against a soft material. cMedium
hardness mudstone piece (MWO_052 from Layer 4, square W1,
2017 excavations) with grinding striations on two surfaces. dA
hard, specular iron oxide nodule (MWO_141 from spit 1224,
square Eg, Tobias excavations) that has two adjacent ground
surfaces. eA large, medium hardness, shale nodule (MWO_147
from spit 1224, square Eg, Tobias excavations) with grinding
use-wear on most surfaces of the piece, forming flat, faceted sur-
faces. fA hard, specular iron oxide nodule (MWO_148 from spit
1225, square Dd, Tobias excavations). The ground and rubbed
surfaces of this piece form a crayon-shaped, faceted tip
Archaeol Anthropol Sci
representation of the distribution of Z-scores over space.
Clearly, areas of varying intensity are drawn, which indicates,
at least, three lithic clusters, with a potential extra one between
IV and V.
2017 Lithic collection
The 2017 excavations yielded 1747 lithic pieces over 2 cm in
length and over 15,000 chips (pieces under 2 cm) (Table 5).
Interestingly, there is an inverse correlation between the vol-
ume of sediments excavated and the abundance of artefacts
over 2 cm (Fig. 13). The most remarkable example of this
inverse proportion is observed in units II and IV, the former
being the one with the most litres excavated but the smallest
amount of artefacts recovered, while unit IV follows the op-
posite trend (most artefacts and second least sediment).
Another significant result is that there is not a single sterile
layer, thus contradicting Tobiass conclusion that two of the
layers were completely devoid of archaeological material.
The raw materials knapped at Mwulus Cave include
quartzite, quartz and chert (Table 5). While in previous anal-
yses (i.e. Watts 1998;Sampson1972), hornfels is also men-
tioned, no hornfel blanks were identified in the new collection.
The dominant raw material in all layers is quartzite, likely
originating from the caves surroundings. Quartzite tools
could have been made from raw material directly collected
from the same outcrop where the cave is located, a potential
reason for the occupation of this site. The second raw material
for units I, III and V is chert, whereas in units II and IV, quartz
is more abundant. The quartz sample of unit II is remarkable
as this layer has very few pieces.
Most of the pieces are well preserved, even though a
small amount in each layer present traces of water patina
or alteration. There does not seem to be a correlation
between weathered pieces and spatial distribution. The
category otherfor raw materials in Table 5refers to
pieces with a high degree of weathering, which made
the identification of the raw material type difficult.
Fig. 12 aVertical distribution of lithics in West profile distinguished by layer. bMap of density of lithic material using kriging extrapolation. cHotspot
analysis. dLithic distribution every 10 cm depth, where three clusters are quite evident
Table 2 Number of cells, pieces and study area for the grid
Cells (10 × 10 cm) Pieces Study area
182 1534* 1.51 m
Archaeol Anthropol Sci
Regarding the main technological categories in all the
layers, there is a clear dominance of flakes and flake frag-
ments, and a similar proportion of flakes and blades, blade/
bladelets throughout the sequence (Table 6and Fig. 14).
Although this is preliminary appreciation, it seems that the
blade production is integrated within the same reduction se-
quence as the flakes. The preliminary technological analysis
shows that most of the blanks present centripetal scar re-
movals. Therefore, most of the production seems related to a
Levallois-like type of reduction that will be described more in-
depth in future technological works.
Regarding the retouched pieces, their percentage amongst
the > 2 cm material is extremely low (Table 7). Unit IV has
provided the most retouched material. The retouched pieces
can be grouped into three main clusters: side-scrapers, notches
and denticulates (Table 7; Fig. 15). While unfortunately no
bifacial piece was found during the new excavations (in con-
trast to Tobiass excavation, Fig. 16), it is worth noting the
recovery of a unifacial point in unit I and of denticulated
points in units III, IVand V, which, when these two categories
are lumped together, constitute the third typological group.
Besides these three groups, we should highlight the pres-
ence of triangular blanks, which also seem integrated within
the Levallois-like production of flakes and blades already
mentioned, although this too requires future technological
analyses to be confirmed. It is also worth noting that 60% of
the triangular blanks occurred in unit IV.
Following the spatial analysis presented in the previous
section, if we group the techno-typological characteristics that
we just highlighted in the three clusters of density of lithics (I,
II, III-IV-V), we see some interesting results. Firstly, the trian-
gular blanks (n= 75) as well as the denticulated points (n=3)
appear mainly in the lower unit (III-IV-V) (Table 7), whereas
unifacial points with simple retouch seem related to the upper
unit I (n= 3). Secondly, regarding the raw material distribu-
tion, it is interesting to note that units I and III-IV-V (lumped
together) have a very similar percentage of rock types repre-
sented, whereas unit II differs with a higher percentage of
quartz and rocks from the category other, which might be
related to post-depositional processes (Fig. 17).
Nonetheless, with this preliminary study, it is still not clear
whether the Levallois technology is consistent across these
three units. Moreover, we have to clarify how flakes, blades
and triangular blanks are related in these three units. Such
techno-functional questions will be the focus of a future pub-
lication specifically dedicated to the lithic technology from
Tobiass lithic collection
Tobias attributed all the archaeological material from his ex-
cavation to the Pietersburg industry (Fig. 17). He also pro-
posed an evolutionary interpretation of the archaeological ma-
terial stating that: From the earlier to the later levels of occu-
pation there is evidence of a progressively more careful selec-
tion of materials.This statement concerns the raw material
representation and the retouched material. Tobias highlighted
in his paper the most remarkable retouched pieces from his
excavation. He paid special attention to some pieces from the
middle level(which we assume is Ash and Sand II), where
he found bifacial and unifacial pieces (Fig. 4; this layer would
tentatively correspond to our unit III). In the upper layer (Ash
and Sand III), he also uncovered unifacial and bifacial pieces;
he mentioned that these layers had more retouched pieces than
the lower ones.
Sampson (1972) listed quartzite, quartz, chert and lydianite
(hornfels) as raw materials. He allocated the trimmed points
to Beds II and III. He also indicated the presence of adjacent
platform cores and discoid cores in Bed I, whereas in unit III
four Levallois cores were documented. As explained above
and see Supplementary Material and Table S1,heconsidered
that Bed II was dubious because of the possibility of admix-
ture during Tobiass excavation. Nonetheless, following his
figure with the stratigraphical reconstruction and correspon-
dence with the Origins Centre material, it seems that all of
Tob iass Beds actually comprised mixed material because of
Table 3 Statistical results for the I Moran index
Spacial autocorrelation
I Moran index Expected Index Variance z-score pvalue Conceptualization Distance method Distance threshold
0.649604 0.005525 0.002889 12.187662 0.000000 Fixed distance = 0.10001 m Euclidean 0.10001 m
Table 4 Nearest neighbour result
Nearest neighbour
Nearest neighbour ratio Observed mean distance (m) Expected mean distance (m) z-score pvalue Distance method
0.787386 0.012683 0.016107 15.930720 0.000000 Euclidean
Archaeol Anthropol Sci
excavation inexperience, rather than stratigraphic problems.
Moreover, as already pointed out, Beds III and II share some
spits, notably spits 1224 and 1225 (Supplementary Material
and Table S1).
Even if spatially we do not know where Tobiasssquares
were situated, we conducted a preliminary evaluation of the
material stored in the Origins Centre, technologically classi-
fying the material from the two squares with a presumably
larger excavation depth (Table S2). Some of the lower- and
uppermost spits (included in Beds I and III respectively) could
be related to units I and III-IV-Vof our cluster lithic analysis.
From the preliminary technological analysis of his beddivi-
sion, it seems that he selected the material, as can be seen in
Fig. 17 where the central spits have a remarkably low propor-
tion of quartz, which does not match with the raw material
distribution in our three clusters (cf. Figs. 17 and 18).
Our renewed excavations at Mwulus Cave have allowed usto
propose a new stratigraphy of the archaeological deposits,
based on sedimentology and the spatial analysis of lithic arte-
fact densities. Furthermore, data on the vegetation surround-
ing the site, based on the pollen and phytoliths preserved in
those deposits, are now available. Such data will be temporal-
ly constrained by luminescence ages at a later stage in order to
provide a more accurate understanding of the
palaeoenvironmental conditions existing during periods of hu-
man occupation of the site. Finally, the cultural archaeological
remains have confirmed the management of ochre pieces
throughout the sequence and the presence of a techno-
typological lithic sequence comprising unifacial pieces and
triangular blanks.
Sedimentology and new stratigraphy
The sedimentology generally fits the deposit accumulation
model for a small cave gradually filled with a combination
of aeolian and autogenic sands with biological contributions
of finer sediments through ash, faecal matter and dust, and
anthropogenic contributions of lithics of varying sizes (bones
are not preserved in the low pH quartzitic autogenic sedi-
ments). Fluid interaction is limited to diffuse vertical drip
Table 5 Raw material
representation for lithic pieces per
Pieces > 2 cm Chips
Quartzite Quartz Chert Other Total per layer
I 217 47 58 4 326 2504
II 66 27 8 11 112 767
III 353 57 95 20 525 3707
IV 444 102 85 12 643 6524
V 107 15 19 0 141 2172
Total per raw material 1187 248 265 47 1747
Fig. 13 Comparison between
volume of sediments (data
presented here in litre
percentages) and lithic material
density (pieces over 2 cm) per
stratigraphic unit
Archaeol Anthropol Sci
water away from the entrance and limited runoff into the cave
from rain and mist (Maguire 2009). Morphologically, a shal-
low gradient is seen sloping into the cave from the dripline and
in this area, low-energy water runoff has played a part in the
redistribution of sediments and biological material away from
the cave mouth into the cave, aiding in the catchment of small-
er anthropogenic material. Where previous excavation walls
were exposed by new excavations, a combination of collapse
and low-energy surface runoff eroded the remnant walls
especially noticeable in the northern wall of the excavation,
closest to the shelter entrance. The limited size, morphology,
location (away from any source of high-energy fluvial or col-
luvial processes) and lithology of the cave restricts any signif-
icant processes of erosion removing deposits. Although depo-
sitional hiatuses and minor erosions are undoubtedly present
and represent potentially significant periods of missing time
in the Mwulus sequence, catchment of archaeological mate-
rial is good and post-depositional movement of larger interred
lithics is limited.
The close association between artefact and clast abun-
dance, moisture and organic matter (and by association sedi-
ment colour) and their noticeable fluctuations through the de-
posit in contrast to an absence of clear changes in unit sedi-
mentological characteristics suggests that Tobiass(1949,
1954) stratigraphic definitions are sedimentologically
unsupported. Structurally, a distinction can be made between
units I and II but it is proposed here that moisture, organic
carbon and colour (light and dark) are post-depositional accu-
mulations favouring artefact-rich layers through the sequence
in a generally well-draining consistently sandy-loam matrix.
In numerous instances through the sequence, modern and old
roots were found directly associated with artefacts and clasts.
Moisture accumulation in artefact dense layers encourages
organic matter accumulation and silt capture with silt quanti-
ties enriched in upper unit III samples compared to units I and
II). Tangible differentiation of units can be made by artefact
abundance, which does distinctly fluctuate through the se-
quence without evidence of significant vertical or lateral
It is posited here that Tobiass stratigraphic sequence, al-
though identifying the same basic units, incorrectly attributed
those units to sedimentological and therefore depositional dif-
ferences. It is likely that although some depositional surfaces
remain intact and represent hiatuses on which artefacts and
clasts accumulated (e.g. between units I and II), the strati-
graphic sequence of Mwulus Cave can be simplified into
three major units represented by occupation of the shelter.
We can therefore propose that the Mwulussequencebe
divided up into three archaeological strata following the lithic
spatial analysis. An uppermost layer represented by unit I
(Tobiass A), a middle layer represented by unit II (Tobiass
B) and lower layer represented by units III, IVand V (Tobiass
C, D and E).
Palaeoenvironments at Mwulus Cave
The pollen results from this investigation refute previous sug-
gestions that there is no fossiliferous material preserved in
MwulusCave(Tobias1954). This is important, should these
records be found to be scientifically valid, as this would
Table 6 Main technological categories per layer
Flakes + flake fragments 271 67 420 504 116
Blade + blade fragments 35 11 56 84 18
Cores 4 1 11 12 4
Retouched 7 3 8 27 2
Chunks 6 30 30 16 1
Total 326 112 525 643 141
Fig. 14 Percentage of flakes and
Archaeol Anthropol Sci
facilitate a more detailed environmental reconstruction than
was afforded by the sedimentary properties alone (Tobias
1949). However, the age and origin of the pollen identified
from these samples must be critically interrogated, particularly
as cave environments prohibit the in situ deposition of pollen,
and consequently, such records are indicative of the broader
regional environment at best (Carrión and Scott 1999). This is
of course true for any pollen record, due to the inherent mech-
anism of wind transport as a method of cross pollination,
facilitating transport distances of at least a 100-km radius
(Andrews and Bamford 2008). However, where pollinating
plants are found in situ, the regional and local influence can
be at least broadly inferred through what is termed the
witches hathypothesis, which considers the proportion of
local grains in any given sample, while the Neves effect
Table 7 Retouched pieces per layer
Retouched flake 1 1 2
Retouched blade 1 1 1 3
End-scraper 1 1
Side-scraper 3 3 6
Indeterminate retouched piece 3 1 9 1 14
Denticulate 2 1 2 5
Unifacial point (simple retouch continuous) 3 1 4
Unifacial point (denticulated retouch) 1 1 1 3
Notch 4 1 5
Total 8 4 6 21 4 43
Fig. 15 Retouched pieces retrieved during the 2017 excavation. a,b. Unifacial points. c,f,i. Triangular blanks. d,e. Denticulated point. g,h.
Denticulates. j. Side-scraper
Archaeol Anthropol Sci
Fig. 16 a,b,f,g. Bifacial and unifacial pieces from Tobiass collection. c,d,e,h,i,j. Triangular blanks from Bed Ifrom Tobiass collection
Archaeol Anthropol Sci
incorporates the role of the size, and thus weight, of different
pollen grains which in turn influences the proportional repre-
sentation of lighter pollen grains within a profile (Prentice
1985; Traverse 2007). For cave environments, further compli-
cations arise from the angle of the cave opening relative to the
angle of dominant and secondary wind directions. In the case
of open excavations, the potential for contemporary pollen
contamination exists (Carrión and Scott 1999), and we there-
fore interpret the pollen diagram for its palaeoenvironmental
significance. Large proportions of grass pollen can be indica-
tive of contamination during excavation and sampling. A
clearer indicator is, however, in the form of exotic species.
Pinus pollen was identified through the profile and is gener-
ally taken as indicative of modern contamination as the spe-
cies was introduced in South Africa ~ 300 years ago (Turner
and Plater 2004; Van Wilgen and Richardson 2012). However,
samples IIe and IIIa contained high pollen concentrations, and
this is not seen in other layers where Pinus is also present. This
suggests that modern contamination does not explain the
higher pollen counts in these two samples, but rather it might
indicate a higher plant cover during these occupation layers.
The relatively low number of pollen grains preserved and the
cosmopolitan nature of their species distributions limit the ca-
pacity for definitive reconstructions of the palaeoenvironment
Fig. 17 Raw materials following
the three lithic clusters proposed
by the spatial analysis
Fig. 18 Lithic raw material
distribution of Tobiassspitsin
square De
Archaeol Anthropol Sci
or palaeoclimate. However, broad inferences can be tentatively
made on the basis of the broad assemblage of the pollen, and the
variations in preservation. The ratio of Poaceae, Cyperaceae
and Asteraceae fossil pollen grains is often used to interpret
the amount and seasonality of maximum regional moisture
availability, and in turn the precipitation (Fitchett and
Bamford 2017).
With regard to the phytolith analysis, the abundance of
phytoliths decreases accordingly with that of other bio-
siliceous microfossils (i.e. diatoms and sponge spicules),
which could be indicative of differential preservation condi-
tions at the site.
Asteraceae, Poaceae, Cyperaceae and Combretaceae fami-
lies dominate the pollen diagram at Mwulus Cave. In unit III,
Cyperaceae pollen grains represent the greatest percentage
composition of sample IIIa, and equal proportions of Poaceae
and Asteraceae grains have been identified. This would suggest
greater regional moisture availability, and a heightened occur-
rence of winter rainfall (Fitchett and Bamford 2017). For the
remaining samples of unit III, insufficient pollengrainshave
been identified to facilitate environmental inferences.
During the occupation of unit II, the two samples with the
greatest fossil pollen concentration, IIe and IIf, grains from the
Poaceae family represented the greatest percentage contribution
of each sample. This was followed by Cyperaceae, with low
percentage composition from the Asteraceae family for both
samples. The phytolith assemblage of unit II in general and of
sample IIe in particular was characterised by the dominance of
grass phytoliths, mainly GSSCs lobates, which are characteris-
tic of the C
Panicoideae subfamily. Sedge phytoliths (papillae/
hat shape and achene morphotypes), blocky parallelepipeds and
elongates without decorated sides were also well represented in
the phytolith assemblage. The pollen and phytolith records ten-
tatively indicate a change in the environmental conditions to-
wards an increase in warmer and drier conditions and the ex-
pansion of open environments where grasses would dominate
the landscapes probably typified by edaphic grasslands.
Summer rainfall conditions wouldbeexpectedduringtheoc-
cupation of unit II. The higher presence of sedges in samples
from unit II might also indicate that local wet conditions were
also present in the surrounding areas of the site.
Based on the phytolith record, the vegetation during the
occupation of unit I experienced an important change
expressed at Mwulus Cave in the form of decreasing grasses,
which, when present, were mostly of the C
type. This,
coupled with a shift in the ratio of Poaceae:Asteraceae pollen
grains, tentatively suggests a return to colder conditions and
possibly a higher occurrence of winter rainfall.
The presence of algal material (including diatoms and
sponge spicules) and spores throughout the record may be of
palaeoenvironmental significance, as this does likely reveal in
situ deposition (Balme 1995). Tobias (1949:p.4)reflectsthat
the deposit was damp throughout, particularly towards the
back of the cave. The water which had gained access by seep-
age had leeched out almost all bony remains…’ The spores
and algal material throughout the record indicate that damp
conditions and moisture availability characterised the cave
throughout the period represented by this profile.
The future availability of ages for MwulusCavewillpro-
vide the opportunity to chronologically contextualise and re-
fine this preliminary palaeoenvironmental reconstruction.
Cultural remains
Archaeologically, Mwulus Cave has been fully attributed to
the Pietersburg industry and compared to three major archae-
ological sequences in the northeastern part of South Africa,
namely Cave of Hearths (Mason, 1988), Border Cave (Tobias
1949; Beaumont 1978) and Bushman Rock Shelter (Mason
Investigating the variability of this presumed industry can
help in tackling some of the issues that we pointed out in the
introduction, such as the idiosyncrasy of the early MSA in the
northeastern area of South Africa, the potential link with other
northern cultural areas, as well as the hypothesis of regional-
ization and development of the MSA in southern Africa.
The Pietersburg was first mentioned by E.G. Paterson to
describe lithic artefacts collected at a mission station near the
town of Pietersburg (today Polokwane) in the late 1920s
(Sampson 1974). Later, Goodwin and van Riet Lowe (1929)
used this term to refer to different lithic surface scatters, which
could be found as far as Victoria West in the Northern Cape. In
1940, van Riet Lowe suggested that it should be elevated to the
category of industry, probably because of the relevance, in his
views, of the Makapan Valley sites (Sampson 1972). Tobias
(1949,1954) assigned the complete lithic assemblage from
Mwulus Cave to the same industry. Finally, Mason (1957)also
identified the Pietersburg in the nearby site of Cave of Hearths,
and, based on his analysis of the lithics from the site, subdivided
it into a Lower (Bed 4), a Middle (Bed 5) and an Upper (Bed 6
9) phase. Sampson (1974) later grouped different southern
African industries under the term Pietersburg complex, name-
ly the Pietersburg industry described by Goodwin, Tobias and
Mason (with Cave of Hearths in the Makapan Valley as the
reference site), the Orangian industryand the Mossel Bay
industry. Following Sampsons description, the Pietersburgs
earlier phase is characterised by utilised flakes, discoid or ad-
jacent cores(sic), large flakes and blades and some frontal
scrapers or backed knives. The later phase, still following
Sampson, has a lower percentage of large blades and higher
percentage of convergent flakes, sometimes converted into
frontal scrapers (Sampson 1974:p.159).
The term Pietersburghas recently been used again (see for
example Lombard et al. 2012;Porrazetal.2015,2018). Porraz
et al. (2015,2018) recognise two techno-typological phases at
Bushman Rock Shelter (Phase 21 and Phase 28). The most
Archaeol Anthropol Sci
recent one is characterised by unifacial and bifacial pieces
(Porraz et al. 2015), whereas the second oneshallmarkisthe
occurrence of abundant end-scrapers (Porraz et al. 2018). Both
of these phases include a strong Levallois component.
One of the main questions surrounding the so-called
Pietersburg relates indeed to its definition. As is often the case
with the MSA, we are dealing here with quite an elusive con-
cept, and there are several reasons for this. The first reason is a
change of status, from a local variation to an industry, and, more
recently, to a technocomplex. Since the 1960s, very few sites in
this part of southern Africa where it was first defined have been
excavated. Moreover, the initial definition of the Pietersburg
was solely based on typological grounds and, as a result, clear
and distinct technological features are lacking. Recent studies
are proposing new technological descriptions but those are ei-
ther still preliminary (Porraz et al. 2015,2018;Backwelletal.
2018) or facing taphonomic issues (Wadley et al. 2016;dela
Peña and Witelson 2018); further work is thus required.
Secondly, the coherence of the Pietersburg industry
through time poses a problem, from a technological point of
view, since its original definition was somewhat vague. As
mentioned above, Mason defined three phases using the
Cave of Hearthssequence, while later on, Sampson
recognised only two phases. Whether the Pietersburg includes
two or three phases, it remains unclear which features are truly
idiosyncratic of this industry.
Thirdly, its exact position within the MSA chronology is
problematic. Most Pietersburg sites were radiocarbon-dated
and have ages that fall beyond the range of the method (>
40 ka BP). The only existing ages using more modern dating
techniques (namely ESR) are for Border Cave, where the
Pietersburg phase has been dated to MIS5 (13080 ka)
(Grün and Beaumont 2001;Grünetal.2003). The attribution
of the base of the sequence to the Pietersburg was initially
proposed by Cooke et al. (1945), followed 30 years later by
Beaumont (1978), but this attribution rests mostly on chrono-
logical grounds rather than thorough lithic analyses.
A further issue surrounding the Pietersburg is its relationship
with other MIS 5 industrial entities, for instance the so-called
MSA I and II from Klasies River Mouth. Indeed, tackling its
definition and relationship with other MIS 5 industries leads us
to another relevant debate, namely the question of whether there
is a regionalization in terms of technological traditions going as
far back as the MIS 6/5 in southern Africa (as first proposed by
Clark 1959). Nonetheless, as mentioned before, it would be
more pertinent to compare MwulusCaves assemblage with
Zimbabwean, Mozambican and Botswanan MSA assemblages,
which are geographically much closer (see for example recent
work in Mozambique by Bicho et al. 2018).
Finally, most of the sites attributed to the so-called
Pietersburg at the beginning of the twentieth century were
open-air sites, and their relationship with cave and rock shelter
sites excavated later requires further clarification.
Mason correlates Mwulus Cave stratigraphy with layers 6
and 7 of Cave of Hearths (Mason 1988). Sampson also studied
and compared the lithic material from Cave of Hearths with
the one from MwulusCave(Sampson1972). He concludes
that, because of Tobiass excavation method problems, the
collection from Mwulus Cave Bed II should be considered
as dubious in any comparative analysis, since there is a high
probability of admixture (Supplementary Material). Sampson
also points out that the broad techno-typological differences
that he describes at Mwulus Cave seem to correspond with
the differences observed between Cave of Hearths Bed 5 and
Beds 68. Cave of HearthsBed 5 was mainly characterised
by convergent flakes, whereas the main hallmarks in Beds 68
are trimmed pieces(a term which refers to unifacial and
bifacial pieces). In our new excavation, unifacial pieces have
been documented in unit I, whereas units IV and V are partic-
ularly rich in triangular blanks. Unfortunately, no dates are
available for Cave of Hearths, which prevents us from draw-
ing chronological comparisons between the two sites.
Regarding Border Cave, a comparison with MwulusCave
is provided by Beaumont in his Mastersthesis(Beaumont
1978). Border Cave, as explained by Tobias, was the first sealed
cave sequence excavated to be associated with the so-called
Pietersburg (Cooke et al. 1945;Tobias1949, p.1). In the old
collection, recovered during Beaumonts excavation, bifacial
and unifacial pieces, as well as triangular elements, are de-
scribed in the basal members. The new excavations started in
2015 (Backwell et al. 2018) have yielded a noticeable amount
of triangular blanks in members 4WA and 5BS, but no bifacial
or unifacial pieces so far. Ages between 227 and 77 kya have
been proposed for members 5WA, 5BS, 4WA and 4BS, all
attributed to the Pietersburg/MSA1 (Grün and Beaumont
2001;Grünetal.2003). Once again, it is worth noting that
the main typological categories found at Mwulus Cave are also
mentioned for Border Cave: unifacial and bifacial pieces and
mainly triangular blanks, being recovered in the upper layers.
The recent publications on Bushman Rock Shelter (Porraz
et al. 2015,2018) highlight three main technological horizons
regarding the lithics. The upper part of the sequence includes
several layers containing bifacial and unifacial pieces (phase
21), while the middle part is characterised by end-scrapers
(phase 28), and the base by a proliferation of triangular
blanks. Recent luminescence dating presented by Porraz
et al. (2018) provides ages for the upper two phases of, re-
spectively, 73 ± 6 ka (quartz) and 91 ± 10 ka (feldspar), and
75 ± 6 ka (quartz) and 97 ± 10 ka (feldspar), thus positioning
this industry within MIS5.
Other sites in Limpopo recently attributed or tentatively
related to the Pietersburg include Wonderkrater (Backwell
et al. 2014) and Steenbokfontein (Wadley et al., 2016).
Wonderkrater has yielded three small MSA lithic assemblages
with age estimates of 30 ka, > 45 ka and 138.01± 7.7 ka BP.
At Steenbokfontein (Wadley et al. 2016), two main knapping
Archaeol Anthropol Sci
methods, namely prismatic blade production and centripetal
flake production, were documented. In Layer 1, there seems to
be different typometric size objectives for flakes and blades,
whereas in Layer 2, a normal distribution for flakes and blades
was detected, which probably indicates a continuous reduc-
tion process. The high proportion of cortex and the small
number of scars on flake dorsal faces might be pointing to-
wards the use of the site as a workshop, perhaps for testing
rock slabs. Steenbokfontein cannot readily be compared to
other MSA sites in Limpopo. This is not only due to the small
sample of lithics analysed but mainly because post-
depositional processes significantly damaged the lithics (de
la Peña and Witelson 2018).
Another sequence that should be compared with the one
from Mwulus Cave is in Rose Cottage Cave in the Free State,
particularly the basal layers excavated by Malan (Wadley and
Harper 1989). Malan attributed the whole sequence to the
Magosian, but the analysis conducted by Wadley and Harper
demonstrated that there were pre- and post-Howiesons Poort
industries. Of particular interest is the pre-Howiesons Poort
material. The so-called Lower Magosian described by Malan
contained many Badvanced^Levallois flakes and blades, and
abundant points and sidescrapers. Wadley and Harpersstudy
shows that some of the most abundant typological categories
of the lower layers were pointsas well as unifacial and par-
tially bifacial pieces.
At Florisbad, another site located in the Free State, Kuman
et al. (1999) described a highly retouched form of MSA at 157 ka
BP and a minimally retouched, expedient MSA assemblage from
a series of occupation horizons at 121 ka BP. They mention small
cores and small cores-on-flakes occurring throughout the se-
quence, predominantly on hornfels. Bifaces are absent from the
assemblages recovered during the 1980s excavation.
At Wonderwerk, in the Northern Cape Province, Beaumont
and Vogel (2006) report on MSA artefacts dating to between
220 and 73 ka BP and including prepared cores, blades,
Levallois flakes and unifacial and bifacial points comparable
with MiddleLate Pietersburg assemblages like those from
Cave of Hearths, Beds 58. In-depth analyses of this industry
have, however, not yet been published.
Bifacial MSA technology has also been highlighted in the
pre-Still Bay layers of Sibudu Cave in KwaZulu-Natal
(Wadley 2012;Rotsetal.2017). Nonetheless, detailed tech-
nological analyses of these lower layers are still pending; a
technological and functional analysis of the serrated points has
been recently published, but those are regarded as a regional
expression (Rots et al. 2017).
Thus, following these broad typological descriptions, we
can assume that, from a lithic point of view, there is in the
northeastern part of South Africa a typological succession
from triangular blanks/Levallois-like industries to bifacial/
unifacial pieces/Levallois-like industries, as confirmed by
the preliminary lithic data from Mwulus Cave. We can also
assume that this succession started (at least) in MIS5, follow-
ing the dates from Bushman Rock Shelter and Border Cave.
The challenge is now to test whether this typological distinc-
tion has a technological counterpart. In other words, is this
typological change also accompanied by a technological one
and, mainly, what does it imply in terms of human behav-
iours? Accordingly, the question is to know whether this lithic
typological sequence is a regional expression and how it re-
lates to older lithic expressions, such as the so-called
Fauresmith, as well as to more recent ones, such as the Still
Bay or the Rhodesian Still Bay. In other words, what is its
diachronic development?
If the chronological program that we have designed to
compare TL and OSL dating confirms an MIS5 chronology,
Mwulus Cave would also be confirmed as a sequence with a
rather old MSA ochre assemblage, as suggested by Tobias
(1949,1954,2005). Nonetheless, the 2017 ochre assemblage
differs from many of the other well-studied MSA ochre as-
semblages, such as those from Blombos Cave, Diepkloof
Rock Shelter and Sibudu Cave. Grinding of ochre pieces is
not directly evident in the 2017 MwulusCaveassemblage,
while this is the most common method of ochre processing at
many MSA sites. There is also no apparent preference to col-
lect red varieties such as is common at many MSA sites, but
100% of the utilised pieces are bright red. The differences
between the two ochre assemblages (Tobiass collection and
2017 collection) at Mwulus Cave are extreme, but excavation
methods could account for some of this. Most of the pieces in
the 2017 assemblage are small, and if excavation and sorting
techniques had not been strict, most of the ochre assemblage
would have been lost. The small piece size and the preference
for soft, clayey-silty ochre suggest that crushing might have
been the favoured way to produce powder. The few hard
pieces (often specular iron oxides; rare in the assemblage)
would have been difficult to crush, and these were used mostly
for grinding and rubbing activities for powder production and
possible direct transferal of ochre powder onto a soft surface.
Some examples of MIS5 ochre use are Pinnacle Point Cave
13B and Klasier River Mouth. Over 500 pieces of ochre were
found at Pinnacle Point Cave 13B from MSA layers dated to
~11592 ka BP (Marean et al. 2007;Watts2010). The utilised
pieces comprise 12.7% of the assemblage, most of which are
red and have been ground (Watts 2010). The MSA ochre
collection from Klasies River Mouth consists of 314 pieces
dated to 11060 ka BP (Singer and Wymer 1982). Most pieces
are red. Utilised pieces were mostly ground and a few have
scored incisions (McBrearty and Brooks 2000;dErrico et al.
2012). By ~ 100 ka, ochre collection and use was habitual in
South Africa (e.g. Watts 2002). Nonetheless, the collection of
ochre in Africa predates these lines of evidence. Ochre pieces
are found at GnJh-15, Kenya (285 kya) (McBrearty 2001)and
in the 250,000-year-old Lupemban Industry at Twin Rivers,
Zambia (Barham 2002). At Sai Island, North Africa, ochre
Archaeol Anthropol Sci
was collected and used by 200 kya in the Sangoan (van Peer
et al. 2004) and ochre was collected at Kathu Pan, South
Africa, in the Fauresmith industry (Wilkins and Chazan 2012).
Renewed multi-disciplinary work at MwulusCaveispartofa
general ongoing trend within the scientific community to in-
vestigate the chronological, technological and cultural charac-
teristics of the MSA from the interior of southern Africa. In
our case, we are particularly concerned with the so-called
Pietersburg industry, to which the lithic material excavated
from the site by P.V. Tobias was assigned. Our reappraisal of
the stratigraphy allows us to propose a model for the site
formation somewhat different from earlier work (Tobias
1949). The sedimentological study combined with spatial
analyses of lithic artefacts highlight a more complex picture
than a direct correspondence between visible sedimentary
units (those identified by Tobias, which are consistent with
ours) and a succession of human occupational events
interrupted by abandonment phases. Rather, while three
phases of site use consistent with lithic concentrations have
been identified (contra five sedimentary units), most of the
sediment accumulation can be explained by the action of
non-anthropogenic biotic (e.g. roots, rootlets and small ani-
mals visiting the cave) and abiotic (i.e. natural erosion of the
cave, autogenic decay of clasts, water percolation and aeolian
inputs) depositional and post-depositional processes.
Although the geological location of the cave, formed within
a quartzitic formation, has prevented bone preservation, re-
mains in the form of pollens and phytoliths are present.
Notwithstanding a possible role played in some instances by
modern contamination and varying degrees of preservation
across the sequence, they provide, for the first time, some in-
formation regarding the palaeoenvironmental setting of the site.
Specifically, the botanical remains indicate climatic fluctuations
through time in the form of changing precipitation patterns.
The results presented here serve as groundwork for future
detailed analyses of both the botanical and cultural remains as
they provide a more secure stratigraphic framework.
Preliminary analysis of the lithic artefacts from the site iden-
tifies techno-typological similarities with material from other
sites in the northeastern part of South Africa, including Cave
of Hearths, Bushman Rock Shelter and Border Cave, thus
tentatively attributing the sequence from Mwulus Cave to
MIS 5/6. Luminescence dating is underway and should help
in refining the chronological attribution of the sequence.
Acknowledgments The authors would like to thank Judy Maguire, Gary
Trower, Lucinda Backwell, Deon Richter and Moloko Madibana, who
helped to locate the site, as well as Chris Thornhill, who helped during the
two excavation campaigns. J.M. Maíllo, María del Carmen Arriaza
Dorado and Lyn Wadley kindly read and gave advice for an early draft
of this manuscript. We thank the Shamane Magashula community for
kindly giving us permission to excavate in MwulusCave.Thesupport
of the DST/NRF Centre of Excellence in Palaeosciences (CoE in
Palaeosciences) towards this research is hereby acknowledged.
Opinions expressed and conclusions arrived at are those of the authors
and are not necessarily to be attributed to the CoE in Palaeosciences. We
are grateful to The Palaeontological Scientific Trust (PAST),
Johannesburg, South Africa and CoE in Palaeosciences for providing
three research grants for MwulusCaveproject.
Albert RM (2000) Study of ash layers through phytolith analyses from the
middle Palaeolithic levels of Kebara and Tabun caves. PhD
Dissertation, Universitat de Barcelona.
Albert RM, Weiner S (2001) Study of phytoliths in prehistoric ash layers
from Kebara and Tabun Caves using a quantitative approach. In:
Meunier JD, Colin F (eds) Phytoliths: applications in earth sciences
and human history, Lisse, pp 251266
Allmendinger RW, Cardozo NC, Fisher D (2013) Structural geology
algorithms: vectors and tensors. Cambridge, England. Cambridge
University Press
Andrews P, Bamford M (2008) Past and present vegetation ecology of
Laetoli, Tanzania. J Hum Evol 54:7898
Backwell LR, Parkinson AH, Roberts EM, d'Errico F, Huchet JB (2012)
Criteria for identifying bone modification by termites in the fossil
record. Palaeogeogr Palaeoclimatol Palaeoecol 337:7287
Backwell LR, McCarthy TS, Wadley L, Henderson Z, Steininger CM,
Barré M, Lamothe M, Chase BM, Woodborne S, Susino GJ,
Bamford MK (2014) Multiproxy record of late Quaternary climate
change and middle stone age human occupation at Wonderkrater,
South Africa. Quat Sci Rev 99:4259
Backwell LR, dErrico F, Banks WE, de la Peña P, Sievers C, Stratford D,
Lennox SJ, Wojcieszak M, Bordy E, Bradfield J, Wadley L (2018)
New excavations at Border cave, KwaZulu-Natal, South Africa. J
Field Archaeol.
Balme BE (1995) Fossil in situ spores and pollen grains: an annotated
catalogue. Rev Palaeobot Palynol 87:81323
Barham LS (2002) Systematic pigment use in the middle Pleistocene of
south-Central Africa. Curr Anthropol 43(1):181190
Beaumont PB (1978) BBorder Cave.^M.A. thesis, University of
Cape Town
Beaumont PB, Vogel JC (2006) On a timescale for the past million years
of human history in Central South Africa. S Afr J Sci 102:217228
Bertran P, Texier JP (1995) Fabric analysis: application to Palaeolithic
sites. J Archaeol Sci 22:521535
Bicho N, Cascalheira J, Haws J, Gonçalves C (2018) Middle Stone Age
technologies in Mozambique: a preliminary study of the Niassa and
Massingir regions. J Afr Archaeol 16(1):6082
Button A (1973) The depositional history of the Welkberg Proto-Basin,
Transvaal. S Afr J Geol 76(1):1525
Button A (1986) The Transvaal sub-basin of the Transvaal sequence.
Mineral deposits of southern Africa, Geological Society of South
Africa, Johannesburg 1:811817
Cabanes D, Shahack-Gross R (2015) Understandingfossil phytolithpres-
ervation: the role of partial dissolution in paleoecology and archae-
ology. PLoS One 10:e0125532.
Cabanes D, Weiner S, Shahack-Gross R (2011) Stability of phytoliths in
the archaeological record: a dissolution study of modern and fossil
phytoliths. J Archaeol Sci 38:24802490.
Archaeol Anthropol Sci
Carrión JS, Scott L (1999) The challenge of pollen analysis in
palaeoenvironmental studies of hominid beds: the record from
Sterkfontein caves. J Hum Evol 36:401408
Clark JD (1959) The prehistory of southern Africa. Harmondsworth, England
Collura LV, Neumann K (2017) Wood and bark phytoliths of west
African woody plants. Quat Int 434:142159.
Conard NJ, Porraz G (2015) Revising models for the cultural stratigraphic
sequence of the Middle Stone Age. S Afr Archaeol Bull 70(201):
Cooke HBS, Malan BD, Wells LH (1945) 3. Fossil man in the Lebombo
Mountains, South Africa: the Border Cave, Ingwavuma District,
Zululand. Man 45:613
Cordova CE (2013) C3 Poaceae and Restionaceae phytoliths as potential
proxies for reconstructing winter rainfall in South Africa. Quat Int
Cordova CE, Scott L (2010) The potential of Poaceae, Cyperaceae and
Restionaceae phytoliths to reflect past environmental conditions in
South Africa. In: Runge J (ed) African Palaeoenvironments and
geomorphic landscape evolution. Taylor & Francis, pp 107134
dErrico F, García Moreno R, Rifkin RF (2012) Technological, elemental
and colorimetric analysis of an engraved ochre fragment from the
middle stone age levels of Klasies River cave 1, South Africa. J
Archaeol Sci 39:942952
Dayet L, Texier P-J, Daniel F, Porraz G (2013) Ochre resources from the
middle stone age sequence of Diepkloof rock shelter, Western Cape,
South Africa. J Archaeol Sci 40:34923505
de la Peña P (2015) Refining our understanding of Howiesons Poort lithic
technology: the evidence from Grey rocky layer in Sibudu cave
(KwaZulu-Natal, South Africa). PLoS One 10(12):e0143451
de la Peña, P, Witelson, D (2018) Trampling vs. retouch in a lithic assem-
blage: a case study from a middle stone age site, steenbokfontein
9KR (Limpopo, South Africa). J Field Archaeol 43(7):522537.
Els BG, van den Berg WA, Mayer JJ (1995) The Black Reef Formation in
the western Transvaal: sedimentological and economic aspects, and
significance for basin evolution. Mineral Deposita 30:112123
Eriksson PG, Schweitzer JK, Bosch PJA, Schereiber UM, Van Deventer
JL, Hatton CJ (1993) The Transvaal sequence: an overview. J Afr
Earth Sci (and the Middle East) 16(12):2551
Eriksson PG, Hattingh PJ, Altermann W (1995) An overview of the
geology of the Transvaal sequence and Bushverld complex, South
Africa. Mineral Deposita 30:98111
Eriksson PG, Altermann W, Catuneanu O, Van der Merwe R, Bumby AJ
(2001) Major influences on the evolution of the 2.672.1 Ga
Transvaal basin, Kaapvaal craton. Sediment Geol 141:205231
Esteban I, De Vynck JC, Singels E, Vlok JHJ, Marean CW, Cowling RM,
Fisher EC, Cabanes D, Albert RM (2017a) Modern soil phytolith
assemblages used as proxies for Paleoscape reconstruction on the
south coast of South Africa. Quat Int 434:160179.
Esteban I, Vlok J, Kotina EL, Bamford MK, Cowling RM, Cabanes D,
Albert RM (2017b) Phytoliths in plants from the south coast of the
greater cape floristic region (South Africa). Rev Palaeobot Palynol
Fitchett JM, Bamford MK (2017) The validity of the Asteraceae: Poaceae
fossil pollen ratio in discrimination of the southern African summer-
and winter-rainfall zones. Quat Sci Rev 160:8595
Fitchett JM, Grab SW, Bamford MK, Mackay AW (2017) Late quaterna-
ry research in southern Africa. Transactions of the Royal Society of
South Africa 72(3):280293
Getis A, Ord JK (1992) The analysis of spatial association by use of
distance statistics. Geogr Anal 24(3):189206
Goodwin, AJH, van Riet Lowe CP (1929). The Stone Age Cultures of
South Africa. Annals of the South African Museum, Cape Town, pp
Grün R, Beaumont P (2001) Border cave revisited: a revised ESR chro-
nology. J Hum Evol 40:467482
Grün R, Beaumont PB, Tobias PV, Eggins S (2003) On the age of border
cave 5 human mandible. J Hum Evol 45:155167
Henry G, Master S (2008) Black reef project. Council for Scientific and
Industrial Research (CSIR) and University of the Witwatersrand,
Henry G, Clendenin CW, Charlesworth EG (1990) Depositional facies of
the black reef quartzite formation in the eastern Transvaal. Extended
Abstracts, 23rd Earth Science Congress (Geocongress' 90) of the
Geological Society of South Africa, Cape Town 90:234237
Henshilwood CS (2012) Late Pleistocene techno-traditions in southern
Africa: a review of the still bay and Howiesons Poort, c. 7559 ka. J
World Prehist 25(34):205237
Henshilwood CS, Sealy J, Yates R, Cruz-Uribe K, Goldberg R,
Grine FE, Klein RG, Poggenpoel C, van Niekerk K, Watts I
(2001) Blombos cave, southern cape, South Africa: preliminary
report on 1992-1999 excavations of the Middle Stone Age
levels. J Archaeol Sci 28:421448
Henshilwood CS, van Niekerk KL, Wurz S, Delagnes A, Armitage SJ,
Rifkin R, Douze K, Keene P, Haaland M, Reynard J, Discamps E,
Mienies S (2014) Klipdrift shelter, southern cape, South Africa:
preliminary report on the Howiesons Poort layers. J Archaeol Sci
Hodgskiss T (2012) An investigation into the properties of the ochre
from Sibudu, KwaZulu-Natal, South Africa. South Afr
Humanit 24:99120
Jacobs Z, Roberts RG (2015) An improved single grain OSL chronology
for the sedimentary deposits from Diepkloof Rockshelter, Western
cape, South Africa. J Archaeol Sci 63:175192
Katz O, Cabanes D, Weiner S, Maeir AM, Boaretto E, Shahack-Gross R
(2010) Rapid phytolith extraction for analysis of phytolith concen-
trations and assemblages during an excavation: an application at Tell
es-Safi/Gath, Israel. J Archaeol Sci 37:15571563.
Key RM (1983) The geology of the area around Gaborone and Lobatse,
Kweneng, Kgatleng, southern and south east districts. District
Memoir Geol Surv Botswana 5:229
Key RM (1986) Sedimentation along the eastern margin of the Bushveld
Basin, SE Botswana. Geocongress '86 (Johannesburg, South
Africa), Abstracts:527530
Kuman K, Inbar M, Clarke RJ (1999) Palaeoenvironments and cultural
sequence of the Florisbad middle stone age hominid site, South
Africa. J Archaeol Sci 26(12):14091425
Lombard M, Wadley L, Deacon J, Wurz S, Parsons I, Mohapi M, Swart J,
Mitchell P (2012) South African and Lesotho Stone Age sequence
updated. S Afr Archaeol Bull 67(195):123144
Maguire J (2009) An overview of the physical setting of Makapan.
In McNabb, J. & Sinclair, A. (eds). The cave of hearths:
Makapan middle Pleistocene research project: field research
by Anthony Sinclair and Patrick Quinney, 19962001. 2948.
Oxford: Archaeopress. University of Southampton Series in
Marean CW (2010) Pinnacle point cave 13B (Western Cape Province,
South Africa) in context: the cape floral kingdom, shellfish, and
modern human origins. J Hum Evol 59(34):425443
Marean CW (2014) The origins and significance of coastal resource use
in Africa and Western Eurasia. J Hum Evol 77:1740
Marean CW, Bar-Matthews M, Bernatchez J, Fisher E, Goldberg P,
Herries AIR, Jacobs Z, Jerardino A, Karkanas P, Minichillo T,
Nilssen PJ, Thompson E, Watts I, Williams HM (2007)Early human
use of marine resources and pigment in South Africa during the
middle Pleistocene. Nat 449:905908
Mason RJ (1957) The Transvaal Middle Stone Age and statistical analy-
sis. S Afr Archaeol Bull 12(48):119137
Archaeol Anthropol Sci
Mason RJ (1988) Cave of Hearths. Makapansgat. Transvaal. Occasional
paper N.21. Archaeological Research Unit
McBreartyS (2001) The middlePleistocene of East Africa. In: Barham L,
McBrearty S, Brooks AS (eds) 2000. The revolution that wasnt: a
new interpretation of the origin of modern human behavior. J Hum
Evol 39: 453563. Robson-Brown (eds), Human roots: Africa and
Asia in the Middle Pleistocene. Bristol: Western Academic &
Specialist Press, pp 8198
McBrearty S, Brooks AS (2000) The revolution that wasnt: a new interpre-
tation of the origin of modern human behavior. J Hum Evol 39:453563
Mucina L, Rutherford MC (2006) The vegetation of South Africa,
Lesotho and Swaziland. South African National Biodiversity
Murungi ML (2017) Phytoliths at Sibudu (South Africa): implicationsfor
vegetation, climate and human occupation during the MSA.
Unpublished PhD thesis, University of the Witwatersrand
Novello A, Bamford MK, Van Wijk Y, Wurz S (2018) Phytoliths in
modern plants and soils from Klasies River, cape region (South
Africa). Quat Int 464:440459.
Partridge TC (1993) The evidence for Cainozoic aridification in southern
Africa. Quat Int 17:105110
Porraz G, Val A, Dayet L, de la Peña P, Douze K, Miller CE, Murungi
ML, Tribolo C, Schmid VC, Sievers C (2015) Bushman Rock
Shelter (Limpopo, South Africa): a perspective from the edge of
the Highveld. South African Archaeol Bull 70:166179
Porraz G, Val A, Tribolo C, Mercier N, de la Peña P, Haaland M, Igreja M,
Miller CE, Schmid V (2018) The MIS5 Pietersburg at 28Bushman
Rock Shelter, Limpopo Province, South Africa. PloS One
Prentice IC (1985) Pollen representation, source area, and basin size:
toward a unified theory of pollen analysis. Quat Res 23(1):7686
Rayner RJ, Moon BP, Masters JC (1993) The Makapansgat australopith-
ecine environment. J Hum Evol 24(3):219231
Repinski P, Holmgren K, Lauritzen SE, Lee-Thorp JA (1999) A late
Holocene climate record from a stalagmite, cold air cave, Northern
Province, South Africa. Palaeogeogr Palaeoclimatol Palaeoecol 150:
Rossouw L (2009) The application of fossil grass-phytolith analysis in the
reconstruction of late Cenozoic environments in the South African
interior. University of the Free State, South Africa
Rots V, Lentfer C, Schmid VC, Porraz G, Conard NJ (2017) Pressure
flaking to serrate bifacial points for the hunt during the MIS5 at
Sibudu cave (South Africa). PLoS One 12(4):e0175151
Rutherford MC, Mucina L, Lötter MC, Bredenkamp GJ, Smit JHL, Scott-
Shaw CR, Hoare DB, Goodman PS, Bezuidenhout H, Scott L, Ellis F,
Powrie LW, Siebert F, Mostert TH, Henning BJ, Venter CE, Camp
KGT, Siebert SJ, Matthews WS, Burrows JE, Dobson L, van Rooyen
Williamson A, Hurter PJH (2006) Savanna Biome. In: Mucina L,
Rutherford M (eds) The Vegetation of South Africa, Lesotho and
Swaziland. South African National Biodiversity Institute, pp 429529
Sampson CG (1972) The Stone Age industries of the Orange River
Scheme and South Africa (No. 6). National Museum Bloemfontein
Sampson CG (1974) The Stone Age archaeology of southern Africa.
Studies in Archaeology, Academic Press
Santisteban Juan I, Mediavilla R, Lopez-Pamo E, Dabrio CJ, Ruiz Zapata
MB, Gil García MJ, Castano S, Martínez-Alfaro PE (2004) Loss on
ignition: a qualitative or quantitative method for organic matter and
carbonate mineral content in sediments? J Paleolimnol 32(3):287299
Schlegel GCJ, Harmse HVM, Brunke O (1989) Granulometric and min-
eralogical characteristics of the Kalahari sands of southern Africa. S
Afr J Geol 92(3):207222
Schwertmann U, Cornell RM (1998) Iron oxides in the laboratory: prep-
aration and characterization. VCH, Cambridge
Singer R, Wymer J (1982) The middle stone age at Klasies River mouth
in South Africa. Chicago University Press, Chicago
Stevenson C, Lee-Thorp JA, Holmgren K (1999) A 3000-year isotopic
record from a stalagmite in cold air cave, Makapansgat Valley,
Northern Province. S Afr J Sci 95:4648
Thackeray JF, Fitchett JM (2016) Rainfall seasonality captured in
micromammalian fauna in Late Quaternary contexts, South Africa.
Palaeontol Afr 51:19
Tobias PV (1949) The excavation of Mwulus cave, Potgietersrust dis-
trict. S Afr Archaeol Bull 4(13):213
Tobias PV (1954) Climatic fluctuations in the middle stone age of South Africa,
as revealed in Mwulus cave. Trans Roy Soc S Afr 34(2):325334
Tobias PV (2005) Into the past: a memoir. Picador Africa, Johannesburg
Traverse A (2007) Paleopalynology. Springer, Dordrecht
Tribolo C, Mercier N, Valladas H, Joron JL, GuibertP, Lefrais Y, Selo M,
Texier PJ, Rigaud JPH, Porraz G, Poggepoel C, Parkington J, Texier
JP, Lenoble A (2009) Thermoluminescence dating of a Stillbay-
Howiesons Poort sequence at Diepkloof rock shelter (Western cape,
South Africa). J Archaeol Sci 36:730739
Tribolo C, Mercier N, Douville E, Joron JL, Reyss JL, Rufer D, Cantin N,
Lefrais Y, Miller CE, Porraz G, Parkington J (2013) OSL and TL
dating of the Middle Stone Age sequence at Diepkloof Rock Shelter
(south Africa): a clarification. J Archaeol Sci 40(9):34013411
Tsartsidou G, Lev-Yadun S, Albert RM, Miller-Rosen A, Efstratiou N,
Weiner S (2007) The phytolith archaeological record: strengths and
weaknesses evaluated based on a quantitative modern reference col-
lection from Greece. J Archaeol Sci 34:12621275.
Turner S, Plater A (2004) Palynological evidence for the origin and de-
velopment of late Holocene wetland sediments: Mdlanzi swamp,
KwaZulu-Natal, South Africa. S Afr J Sci 100:220229
Tyson PD, Preston-Whyte RA (2005) The weather and climate of south-
ern Africa. Oxford University Press, Cape Town, pp 144146
Van Peer P, Rots V, Vroomans JM (2004) A story ofcolourful diggers and
grinders: the Sangoan and Lupemban at site 8-B-11, Sai island,
northern Sudan. Before Farming 3:Article 1
Van Wilgen BW, Richardson DM (2012) Three centuries of managing
introduced conifers in South Africa: benefits, impacts, changing
perceptions and conflict resolution. J Environ Manag 106:5668
Wadley L (2008) The Howieson's poort industry of Sibudu cave.
Goodwin Series 1:122132
Wad ley L (2012) Twomoments in time during middle stone age occupa-
tions of Sibudu, South Africa. South Afr Humanit 24(1):7997
Wadley L (2015) Those marvellous millennia: the Middle Stone Age of
southern Africa. Azania 50:155226
Wadley L, Harper P (1989) Rose cottage cave revisited: Malans Middle
Stone Age collection. S Afr Archaeol Bull 44:2332
Wadley L, Murungi ML, Witelson D, Bolhar R, Bamford M, Sievers C,
Val A, de la Peña P (2016) Steenbokfontein 9KR: a Middle Stone
Age spring site in Limpopo, South Africa. S Afr Archaeol Bull
Watts I (1998) The origin of symbolic culture: the middle stone age of
southern Africa and Khoisan ethnography. Ph.D. Thesis, University
College, London
Watts I (2001) Ochre in the middle stone age of southern Africa: ritualized
display or a hide preservative? S Afr Archaeol Bull 57:114
Watts I (2002) Ochre in the middle stone age of southern Africa: ritualized
display or hide preservative? S Afr Archaeol Bull 57:6474
Watts I (2010) The pigments from pinnacle point cave 13B, Western cape,
South Africa. J Hum Evol 59:392411
Wilkins J, Chazan M (2012) Blade production ~500 thousand years ago
at Kathu Pan 1, South Africa: support for a multiple origins hypoth-
esis for early middle Pleistocene blade technologies. J Archaeol Sci
Will M, Bader G, Conard NJ (2014) Characterizing the Late Pleistocene
MSA lithic technology of Sibudu, KwaZulu-Natal, South Africa.
PLoS One 9:e98359
Archaeol Anthropol Sci
... Recent fieldwork conducted in the Savanna Biome, south of the Limpopo River, constitutes several parallel attempts to compensate for the limited knowledge regarding the nature of MSA developments in the interior of southern Africa. This notably includes fieldwork at Bushman Rock Shelter (Porraz et al. 2015(Porraz et al. , 2018, Mwulu's Cave (de la Peña et al. 2019), Steenbokfontein (Wadley et al. 2016), Wonderkrater (Backwell et al. 2014) and Ga-Mohana Hill North Rockshelter (Wilkins et al. 2020) (Fig. 1). It also includes fieldwork at Border Cave, further east (Backwell et al. 2018) and at Grassridge Rockshelter, today in the Highveld Grassland but close to the limit of the Sub-Escarpment Savanna (Ames et al. 2020). ...
... 18). We refer to Porraz et al. (2018) and de la Peña et al. (2019) for recent historiographic accounts of this industry. Here, we highlight several problematic aspects related to the exact nature of the Pietersburg, which transpire in renewed discussions on this topic (Wadley et al. 2016;Porraz et al. 2018;de la Peña et al. 2019;Chazan et al. 2020;Feathers et al. 2020). ...
... We refer to Porraz et al. (2018) and de la Peña et al. (2019) for recent historiographic accounts of this industry. Here, we highlight several problematic aspects related to the exact nature of the Pietersburg, which transpire in renewed discussions on this topic (Wadley et al. 2016;Porraz et al. 2018;de la Peña et al. 2019;Chazan et al. 2020;Feathers et al. 2020). ...
Full-text available
Olieboomspoort is one of the few rock shelters in the vast interior of southern Africa documenting pulses of occupation from the Acheulean until the end of the Later Stone Age. Revil Mason excavated the site in 1954 and attributed the large Middle Stone Age (MSA) lithic assemblage to his middle phase of the so-called Pietersburg Industry. Recent work at the site has focused on the Holocene layers, but little is known about the earlier phases of shelter use. Here, we provide some background to the shelter, give a history of past research and present initial results following renewed fieldwork at the site. The MSA deposits contain abundant lithic artefacts and ochre, and we present an initial description of these cultural remains. Palynological analysis reveals limited potential for palaeoenvironmental reconstructions, but some faunal remains indicate open grasslands. We dated two equid teeth that provided highly consistent combined U-series-ESR estimates, resulting in a mean age of 150 ± 14 ka (1σ). Even when considering potential sources of uncertainty such as variations in water-intake, these fossils can securely be dated to Marine Isotopic Stage 6. Our reappraisal of site formation processes highlights the fact that the archaeological assemblage is strongly time-averaged. We discuss these different results in the context of a recently rekindled interest in the so-called Pietersburg Industry.
... In 2018, we initiated a geoarchaeological project to study the Pleistocene stratigraphy and depositional history of the site. This fits the trend of recent studies that focus on understanding the depositional history of MSA sites in South Africa (e.g., Backwell et al. 2018;de la Peña et al. 2019;Goldberg et al. 2009;Haaland et al. 2017;Henshilwood et al. 2014;Larbey et al. 2019;Lotter et al. 2016;Miller et al. 2013). Our fieldwork has three interrelated goals. ...
... This may partly account for the gradual and diffuse stratigraphic boundaries. Important in situ weathering input is also seen at other sites, such as Pinnacle Point 5-6 (Karkanas et al. 2015) and Mwulu's Cave (de la Peña et al. 2019). ...
... Sheetwash processes from surrounding slopes also appear absent as the fine bedding produced in such sedimentary regimes was not observed. Aeolian sediments are characteristically well-sorted; their presence would have resulted in the second peak of fine-grained sediments (see de la Peña et al. 2019). The sediments are consistently poorly sorted and negatively skewed (fine tail) and show a single peak (Fig. 6). ...
Full-text available
Umhlatuzana rockshelter has an occupation sequence spanning the last 70,000 years. It is one of the few sites with deposits covering the Middle to Later Stone Age transition (~40,000–30,000 years BP) in southern Africa. Comprehending the site’s depositional history and occupation sequence is thus important for the broader understanding of the development of Homo sapiens’ behavior. The rockshelter was first excavated in the 1980s by Jonathan Kaplan. He suggested that the integrity of the late Middle Stone Age and Later Stone Age sediments was compromised by large-scale sediment movement. In 2018, we initiated a high-resolution geoarchaeological study of the site to clarify the site formation processes. Here, we present the results of the excavation and propose a revised stratigraphic division of the Pleistocene sequence based on field observations, sedimentological (particle size) analyses, and cluster analysis. The taphonomy of the site is assessed through phytolith and geochemical (pH, loss on ignition, stable carbon isotope) analyses. The results indicate a consistent sedimentological environment characterized by in situ weathering. The analysis of the piece-plotted finds demonstrates semihorizontal layering of archaeologically dense zones and more sterile ones. There was no indication of large-scale postdepositional sediment movement. We show that the low-density archaeological horizons in the upper part of the Pleistocene sequence are best explained by the changing patterns of sedimentation rate.
... Although the interior regions of South Africa present an archaeological record less studied than that recovered from the western and southern coasts, the number of sites with MSA deposits is considerable. Some examples come from ongoing or recently excavated sites such as Kathu Pan (Wilkins and Chazan 2012;Lukich et al. 2019Lukich et al. , 2020, Ga-Mohana Hill North Rockshelter (Wilkins et al. 2020) and Wonderwerk in the southern Kalahari Basin; Florisbad (Toffolo et al. 2017) and Baden-Baden (van Aardt et al. 2015) in the Free State or Wonderkrater (Backwell et al. 2014), Bushman Rock Shelter (Porraz et al. 2015), Cave of Hearths (McNabb et al. 2009) and Mwulu's Cave (de la Peña et al. 2019) in Limpopo. ...
... The present study focuses on the sediments from Mwulu's Cave, a Middle Stone Age site geographically situated in the Makapansberg Mountains, within the Limpopo basin, and virtually less than 100 km away of the Kalahari basin ( Fig. 1) (Tobias 1949;de la Peña et al. 2019). The site has been recently dated to~90 ka for all its layers (Feathers et al. 2020, and discussion therein). ...
... The site has been recently dated to~90 ka for all its layers (Feathers et al. 2020, and discussion therein). The deposits contain stone tools and ochre, while faunal remains are absent (de la Peña et al. 2019). Here, we present a study that combines pollen, phytoliths, aquatic bio-silica micro-remains and infrared spectroscopy analysis of Mwulu's Cave archaeological deposits. ...
Full-text available
The interior regions of South Africa have had less attention devoted to archaeological research than coastal regions, and palaeoenvironmental studies are also more limited. As such, little is known about the interaction between human behaviours and past environments in these semi-arid regions. Here, we present an archaeobotanical and mineralogical study from the Middle Stone Age site of Mwulu’s Cave, Limpopo Province. Our study shows the importance of using taphonomical approaches prior to interpreting archaeobotanical assemblages, while provides with novel information on the plants used by ancient inhabitants of Mwulu’s. The grass phytolith composition is of environmental significance, where a shift from C4 Panicoideae to C3 grasses is observed in the last occupation event. This tentatively suggests a shift in rainfall regime, from summer rainfall conditions to an increase in winter rain, during Marine Isotope Stage 5b in the Polokwane region, or a decrease in rainfall seasonality. Although we are unable to chronostratigraphically associate this change in the plant composition, our study adds evidence in support of previous propositions for an expansion of the winter rainfall zone into the interior regions of South Africa.
... The reinvestigation of previously published sites in South Africa is becoming increasingly common as archaeologists seek to apply new technologies and approaches at sites which have yielded significant (or potentially significant) assemblages and sequences, but for which we have relatively limited stratigraphic and contextual information by current standards. Sites with recent or on-going reinvestigations include Border Cave (Backwell et al., 2018), Bushman Rock Shelter (Porraz et al., 2015), Diepkloof Rock Shelter (Porraz et al., 2013), Elands Bay Cave (Porraz et al., 2016), Grassridge Rock Shelter (Ames et al., 2020), Mwulu's Cave (de la Peña et al., 2019), Olieboomspoort (Val et al., 2021) and Umhlatuzana (Sifogeorgaki et al., 2020). Therefore, the issues involved in correlating stratigraphic units (and the assemblages they have yielded) across multiple stratigraphic systems-with differing approaches to excavation and/or recording and varying stratigraphic resolution-is currently highly relevant to MSA archaeological research in the region. ...