STEENBOKFONTEIN 9KR: A MIDDLE STONE AGE SPRING SITE
IN LIMPOPO, SOUTH AFRICA
LYN WADLEY1*, MAY L. MURUNGI1, DAVID WITELSON2,
ROBERT BOLHAR3, MARION BAMFORD1, CHRISTINE SIEVERS2,
AURORE VAL1,4 & PALOMA DE LA PEÑA1,2
1Evolutionary Studies Institute, University of the Witwatersrand, Johannesburg, South Africa
*Corresponding author. E-mail: firstname.lastname@example.org
2Archaeology Department, School of Geography, Archaeology and Environmental Studies,
University of the Witwatersrand, Johannesburg, South Africa
3Geology Department, School of Geosciences, University of the Witwatersrand, Johannesburg, South Africa
4Ditsong National Museum of Natural History, 432 Visagie St, Pretoria, South Africa
(Received April 2016. Revised August 2016)
Steenbokfontein 9KR is a Middle Stone Age spring site in the
Waterberg, Limpopo. It is situated in a geological remnant, the
Vaalwater Formation, a former basin that filled with fine-grained
siltstone and sandstone. The ready supplies of water and siliceous rock
attracted Stone Age settlement. The petrographic and XRF analyses
suggest that the rock used is silicified siltstone. The outcrop is exposed
at the spring and the site appears to have been deliberately exploited for
tool-making rock. Here, people tested rock slabs for their suitability
and knapped some flakes and blades on site. The excavated area shows
signs of post-depositional disturbance, and damage that resembles
trampling is present on both lithics and geological pieces of siltstone.
Phytolith preservation is excellent and a woodland savanna is implied
by the identifications. Cyperaceae once grew around the spring,
although they no longer do.
Key words: Middle Stone Age, spring site, phytoliths, silicified
siltstone, lithic technology.
The Middle Stone Age (MSA) spring mound site, Steenbok-
fontein 9KR (hereafter called Steenbokfontein), is on the farm
of the same name on the Waterberg plateau, Limpopo Province
(Steenbokfontein 9KR (Portion 1) S24.04.661; E28.05.634)
(Figs 1 & 2). Steenbokfontein is unusual in Limpopo, archaeo-
From the 1950s onwards, archaeologists excavating MSA
sites in the interior of South Africa recognised a lithic industry
containing long blades, truncated blades with retouched
edges, and long unifacial points. They named it after the town
of Pietersburg (now Polokwane). Pietersburg Industries are
located principally in the north of South Africa, but they have
not yet been documented north of the Limpopo River. Most
Pietersburg sites in Limpopo Province are caves or rockshelters,
the best known being Cave of Hearths (Mason 1962, 1988;
Sampson 1974; Sinclair 2009), Olieboomspoort (Mason 1962;
Van der Ryst 2006), Bushman Rock Shelter (Plug 1981; Porraz
et al. 2015) and Mwulu’s Cave (Tobias 1949; Sampson 1974). The
open site Blaaubank, a gravel donga near Rooiberg, has many
felsite and quartzite Pietersburg tools overlying Earlier Stone
Age ones (Mason 1962). Another open site, Kalkbank, also
reported to have a Pietersburg industry, yielded only a few
dozen lithics (Mason 1962) amongst the large faunal collection
that is now known to have been accumulated predominantly
by non-human agents (Hutson & Cain 2008). All these sites are
below the Waterberg escarpment and, apart from Kalkbank, all
have extensive sequences potentially beginning in an early
phase of the MSA or even before that in the Earlier Stone Age.
Their lithics are mostly made on rocks locally available in size-
able chunks that enable knapping of large, elongated products.
The Pieterburg Industry, as presently defined, may well be a
response to the availability of hefty blocks of fine-grained rock.
Morphological characteristics are recorded in detail for Cave of
Hearths, but this is not the case for most Pietersburg industries,
and the term has often been used loosely, and clearly needs
revision. It is also necessary to study more open sites in northern
South Africa to see how their assemblages compare with the
better researched ones from cave and rock shelter sites. A
recently-excavated MSA spring and peat site, Wonderkrater
(Barré et al. 2012; Backwell et al. 2014), which is near
Mookgopong (formerly Naboomspruit), has contributed to
this aim. Unfortunately, the late MIS6 and MIS3 occupations
have few lithics. Most are made on rhyolite; broken and whole
flakes are in the majority, followed by broken blades and
denticulates, many of which seem to be cutting tools (Backwell
et al. 2014).
Cave of Hearths (Mason 1962, 1988; Sampson 1974) has a
particularly long Pietersburg sequence. MSA lithics from Beds
5–8 include prepared cores, long blades, Levallois flakes, and
unifacial and bifacial points occasionally on hornfels, but more
often on quartzite and andesite (Mason 1962, 1988; Sampson
1974; Sinclair 2009). Long blades sometimes were manufac-
tured from blocks of locally available andesite (Sinclair 2009).
Olieboomspoort Rock Shelter is another MSA site of consider-
able significance (Mason 1962). Here a Pietersburg Industry
made from felsite, quartzite, cryptocrystalline minerals and
mudstone overlies an Earlier Stone Age (ESA) assemblage and
underlies a long Later Stone Age (LSA) sequence (Mason 1962;
Van der Ryst 2006). Bipolar flaking is part of this Pietersburg
Industry (Mason 1962) as it is at the earlier Limpopo site, Kudu
Koppie (Sumner 2013). Mwulu’s Cave is thought to contain a
late Pietersburg made on quartzite (Tobias 1949; Sampson
1974), as is also the case at Kalkbank (Sampson 1974) and
Blaaubank (Mason 1962). Much farther east, the hornfels-
dominated assemblages from Bushman Rock Shelter have long
blades, retouched blades and elongated unifacial points (Plug
1981; Porraz et al. 2015).
Most excavated MSA sites in Limpopo are below the
escarpment, but amongst the known ones on the Waterberg
plateau, where Steenbokfontein is situated, is a small rock
shelter, North Brabant (New Belgium 608 LR), which was exca-
vated by Schoonraad and Beaumont (1968). Here the MSA
130 South African Archaeological Bulletin 71 (204): 130–145, 2016
South African Archaeological Bulletin 71 (204): 130–145, 2016 131
component of the site was attributed to a ‘Middle Pietersburg’,
and LSA artefacts were recovered from younger occupations
with radiocarbon ages of AD 900 and AD 1100. Van der Ryst
(1998) subsequently excavated the same shelter and retrieved
MSA unifacial points from the oldest sediments. A few MSA
blades, a bifacially worked tool and broken point were later
found in a rock shelter at Schurfpoort 112 KR on the Waterberg
plateau (Van der Ryst 1998). At Goergap 113 KR rock shelter,
also on the plateau, van der Ryst (1998) excavated a substantial
MSA lithic collection. At the base, the lithics were generally
larger than 50 mm and the length of some quartzite blades
exceeded 150 mm. Points and knives (elsewhere sometimes
called side scrapers) were common. In younger MSA layers of
Goergap, the tools were smaller and there was a change in rock
type from quartzite to felsite.
As this brief review demonstrates, few MSA sites have been
FIG. 1. Limpopo Middle Stone Age sites mentioned in the text. Steenbokfontein is marked with an asterisk. BRS, Bushman Rock Shelter; COH, Cave of Hearths;
GG, Goergap; KB, Kalkbank; KK, Kudu Koppie; MC, Mwulu’s Cave; NB, North Brabant; OBP, Olieboomspoort; RB, Rooiberg (Blaubank); SP, Schurfpoort;
FIG. 2.Steenbokfontein spring site with the excavation grid and datum. (A)The spring; (B)the excavation grid and datum.
studied in the Waterberg and Steenbokfontein is the first MSA
open site excavated on the plateau. Since the Waterberg is a
high-lying island in Limpopo, and is encircled by lower-altitude
Pietersburg sites, we want to know whether the different
locations and rock types influence the type of sites and lithics.
Hornfels, for example, does not outcrop on the Waterberg
plateau. The Waterberg is overwhelmingly characterised by
coarse sandstone, yet the Steenbokfontein vicinity has abun-
dant outcrops of fine-grained siliceous rock. This source of
rock, suitable for knapping, attracted people in the MSA, and
the area is littered with the products of knapping. The spring is
an added attraction, so LW selected the site for a trial excava-
tion. It was clear from the start that there would be challenges,
probably more so than is normal for open sites. We did not
expect organic preservation for example, because Waterberg
sedimentsareusuallyacidic.Furthermore,thedensity of lithics
lying on the ground surface points to low sedimentation rates
and repeated visits by people. The water source would also
have enticed animals regularly, so traffic would have been
heavy around the spring. Our pre-excavation investigation of
the surface lithics revealed secondary edge modification on all
the pieces, and we suspected that at least some of this was the
result of repeatedly trampling the siliceous rock. We realised
that a detailed taphonomic study would be required as part of
the lithic analysis.
Steenbokfontein is situated within the Vaalwater Forma-
tion which developed in the central part of the main Waterberg
Basin (Callaghan 1993:51). The Vaalwater Formation is the
youngest part of the Waterberg succession (Jansen 1982). It is
responsible for building the plateau in the area and the farm,
near Visgat, is close to where the Formation reaches its maxi-
mum thickness of approximately 475 m (De Vries 1970: 51). The
Vaalwater beds were most likely deposited under shallow
water in a slowly subsiding, small and isolated inland basin
(Jansen 1982: 53) that was as narrow as 6 km northwest of
Vaalwater and no more than 40 km wide between Vaalwater
and Dorset (De Vries 1970, 1973). Callaghan (1993: 70) interprets
the Vaalwater Formation as the end result of deposits in a
littoral or shallow siliciclastic sea environment. The Formation
has sediments that are finer-grained than those in the underly-
ing Formations, such as the Cleremont Formation, and it
consists of sandstone, siltstone and locally developed gritty
sandstone (De Bruiyn 1971). The sandstone is rich in feldspar,
plagioclase and orthoclase, whereas the siltstone comprises
quartz, plagioclase, orthoclase, zircon and mica (De Bruiyn
1971). In places the siltstone is hardened where diabase intru-
sions caused contact metamorphism (De Bruiyn 1971: 6) and
some siltstones are locally silicified (De Vries 1970; Jansen 1982:
50). Callaghan cut several thin sections for petrographic analy-
sis. Thin section #1577 contained 89% quartz, 10% plagioclase
feldspar, 1% chert and extensive quartz overgrowths, while
thin section #1563 comprised 80% clay (K-rich), 20% silt, iron
oxide cement and grains prolate to equate (Callaghan 1993: 13).
Today, the fine-grained sediments are exposed mostly in
streambeds where they are jointed (Jansen 1982: 51) and have a
blocky appearance. At Steenbokfontein, fine-grained, jointed
siltstones are exposed around the eyes of springs and these are
the rocks that were exploited by people in the MSA. We
thought it was important to conduct a petrographic study of
the rock used for the lithics so that we could identify it securely
and also characterise it because it is unique to this part of South
Africa. Since it occurs in such a small locality it will later be
possible to track rocks exported to other sites.
Several springs rise on the farm and one was selected for
a small excavation. The presence of reliable water sources is
unusual in the Waterberg, notwithstanding its name. The
mean rainfall of the area is approximately 600 mm per annum
(summer rainfall) and the vegetation is predominantly
broad-leaved savanna with sour C4grass. Seasonal drought is
not uncommon, but there are also years of rainfall well above
the mean. Combretum spp., Terminalia sericea,Burkea africana,
Acacia karroo and Grewia spp. are among the common trees and
shrubs, while grasses include Eragrostis spp., Panicum spp.,
Cenchrus ciliaris and Digitaria spp. At present there is no notice-
able difference between plants growing around the spring and
elsewhere in the savanna surrounds and there are no
Cyperaceae near the water.
A north-facing grid was mapped using an EDM (Fig. 2).
Two permanent markers were painted yellow; one is a metal
dropper cemented into the ground with a beacon of rocks,
while the other is a bolt cemented into a concrete slab near the
spring eye. The bolt in the concrete slab is the datum. Two
metre squares (N4 and N5) were excavated to the base of the
surface layer. Thereafter, a half metre (100 × 50 cm) was exca-
vated to bedrock in square N4. All sediment was sieved
through 2 mm mesh and the volumes of deposit removed were
recorded (non-artefactual rock with any dimension larger than
5 cm was removed before the volumes were measured). Sedi-
ment colour was documented using a Munsell chart. Lithics of
2 cm and larger were point plotted. Sediment cores were
removed for Optically Stimulated Luminescence (OSL) dating
fromthebaseoftheSurfacelayer (depth 13 cm), squareN4,east
wall, and the base of Layer 2 (depth 32 cm), square N4, north
wall. Sediment samples were collected from all layers; some
were specifically collected for phytolith analysis.
Four layers were recognised between the surface and bed-
rock that was reached between 52 and 54 cm below surface
(Fig. 3). The surface silt (Surface layer) comprises recent soil
formation (Munsell colour 7.5YR 3/2 very dark greyish-brown;
texture clayey; coherence strong) and the upper 5 cm was
disturbed. Seventy litres of sediment were removed from N4,
and 100 from N5. Twopieces of glass were recovered and lithics
in both squares lay in disarray at various angles. The contact
between Surface and rock-filled Layer 1 is abrupt (Fig. 3) and
excavation became difficult because of the density of rocks.
Layer 1 sediment was a silty texture with medium coherence,
and the Munsell colour was 7.5YR 6/2 pinkish-grey. Seventy-
five litres were excavated from N4 and 40 from N5. Excavation
in N5 was then discontinued. Near the base of Layer 1 the sedi-
ment is gritty and lithic concentrations are dense (Table 1). At
132 South African Archaeological Bulletin 71 (204): 130–145, 2016
TABLE 1.Changing density of Steenbokfontein lithics in square N4 through
Layer nlitres nplotted lithics >2 cm Lithics per litre
Surface 70 128 1.8
Layer 1, top 20 37 1.9
Layer 1, to 22 cm 40 49 1.2
Layer 1, base to 28 cm 15 58 3.9
Layer 2, top to 30 cm 10 20 2.0
Layer 2, to 35 cm 20 21 1.1
Layer 2, base to 40 cm 4 14 3.5
Layer 3, to bedrock 54 15 0.3
South African Archaeological Bulletin 71 (204): 130–145, 2016 133
the base of Layer 1, at a depth of between 28 and 29 cm, there is
a stratum of small rock slabs and this was used as the marker
between Layer 1 and Layer 2. Layer 2 sediment is gritty with
rounded nodules of decomposed sandstone, weak coherence,
and far fewer lithics than in Layer 1. The colour of Layer 2 is
7.5YR 6/2 pinkish-grey and 34 litres of sediment were exca-
vated. The close-packed angular rocks in Layers 1 and 2 are not
in primary context; they lie at various angles and have there-
fore been moved. The outcrop from which they originate lies
below Layer 3 which comprises 54 litres of gritty sediment in
pockets between the many close-packed rocks that form bed-
rock.Therubbleatthebase of Layer 3 formedfromthedesegre-
gation of the siltstone regolith. Only a few lithics were
Twohundred (200) g of sediment from Sur face, and 200 g of
sediment from Layer 1 were processed by flotation by C.S.,
using clear water and chiffon cloth screens with a maximum
mesh size of 500 µm. The carbonised remains that were recovered
through flotation consisted exclusively of grass root fragments,
up to 20 mm long and less than 3 mm in diameter. They occurred
from Surface into Layer 1 at a depth of 25 cm. Some roots were
burned although the farm had not experienced a veld fire for
more than 30 years (J. van der Walt, pers. comm. 2014). There
wasnoorganicpreservationofboneor charcoal in thesediments.
Two sediment cores (one from Surface, and one from Layer
2) were processed by Professor Zenobia Jacobs in the School of
Earth Sciences, University of Wollongong, Australia. Reliable
OSL ages were not possible because grains of different ages are
mixed in the sediments. This implies post-depositional distur-
bance of the area around the spring, a conclusion also sup-
ported by the stratigraphy and the lithic analysis at the site.
Athinsectionwascutfrom a geological block collectednear
the Steenbokfontein spring. The thin section was prepared
following standard techniques, and examined using a trans-
mitted light microscope (Leica 2500 DM P) at the School of
Geosciences, University of Witwatersrand. Petrographic exam-
ination included mineral identification, mineral shape, tex-
tures, and distinction between larger clasts and groundmass.
Classification into rock type was based on petrographic obser-
This sample is a largely homogeneous, unfoliated and fine-
to cryptocrystalline rock consisting of mineral constituents of
variable grain-size (Fig. 4). The fine-grained material can be
considered as matrix composed of cryptocrystalline quartz,
minor carbonate and iron-oxide (hematite), the latter lending
therockitsreddishcolour. Microcrystalline (micritic) carbonate
cannot be specified unambiguously, but is probably dolomitic
(Ca-Mg rich) and/or sideritic (Fe-rich) in composition. Brownish
to reddish hematite is abundant, occurring as granular, finely
disseminated grains. Hematite is also identified as an infilling
FIG.4.Thinsectionphotomicrographs (plane polarisedlight)ofrock sample fromSteenbokfontein.(A)Elongate,angular,randomlyorientated clasts in amatrixof
cryptocrystalline quartz, minorcarbonateandiron-oxide; (B)angular,elongate clast and veinletsinfilledwithquartz and hematite. Carbonaceousmattervisible as
irregularly-shaped clots; angular opaques are probably Fe-oxides; (C)magnified view showing quartz, carbonate matrix and carbonaceous matter and opaques.
FIG. 3.Steenbokfontein south wall stratigraphy of square N4, showing the silty Surface layer, rubble-filled Layers 1–3 and the jointed siltstone bedrock.
phase, replacing primary material in clasts that are elongate,
angular and often needle-like in shape with a random orienta-
tion.Theseclastsrangeinlengthupto 2 mm, and are sometimes
observed to be concentrated along specific horizons. Weathered,
rounded brown clasts of carbonate (dolomite?) are also present.
Opaque phases (~1% modal abundance) are difficult to iden-
tify, but may comprise sulfides and Fe-oxides (magnetite) that
often appear in close association. Rare, detrital quartz grains
show undulose extinction and are distinctly larger than matrix
components. In places, the arrangement of quartz and Fe
oxides resembles pressure-solution textures. Randomly,
veinlets cross-cutting the rock are infilled with quartz and
hematite. At high magnification, carbonaceous matter can be
discerned as irregularly-shaped clots at micron scale. The nature
of the groundmass quartz is not clear, and may represent detrital
or precipitated material. In the former case, the rock would be
classified as silicified siltstone, in the latter case as chert. XRF
analysis has helped to resolve this issue.
This study was carried out using a pXRF: a Thermo Scien-
tific Niton XL3t 950 portable XRF analyser equipped with a
GOLDD+ drift detector and a miniaturised X-ray tube with an
excitation source of 50 kV. An area of 8 mm in diameter was
analysed. Measurements were acquired using a lead receptacle
stand into which the spectrometer is placed. Using this receptacle
and testing the flat surfaces of rocks keeps the distance
between the sample and the spectrometer constant. Acquisi-
tion time for all samples was 180 seconds. Site choice for read-
ing spots is limited by surface topography, because a flat
surface is preferable. pXRF results on whole samples are
inferior to XRF readings on powdered samples. This pXRF test-
ing method was, however, considered adequate for a prelimi-
nary study of the rocks. Two rocks from the Steenbokfontein
spring were sampled and each rock was split so that four pieces
The preliminary XRF readings demonstrate that the four
rock samples are silicified; one sample had Si readings of 31%,
while the other three were between 41 and 46%. They contain
relatively high percentages of aluminium (4–8%), iron (4–5%)
andpotassium (1.6–2.6%). In combination withthe petrographic
analysis, the XRF results suggest that the rocks used for tool
manufacture at Steenbokfontein are silicified siltstone. Although
the rock is cherty in appearance, the relatively low percentages
of silica imply that silicified siltstone is a closer match than
COUNTING AND CLASSIFICATION
Of the sediment samples collected from square N4, four
were chosen for phytolith analysis from Layers 1 to 3 from both
the southwestern and southeastern quadrants. Phytolith
morphotypes were extracted following standard procedures
described in Pearsall (2000) and Piperno (2006). The procedure
involved treating approximately 2–3 g of sediment with 10%
hydrochloric acid in a hot water bath at 70°C to remove carbon-
ates, then washing in distilled water by centrifuging and
decanting. Organic matter was removed from the samples by
adding 30% hydrogen peroxide, and placing in a hot water
bath at 70°C. Samples were sieved through a 250 µm sieve and
finally density separation of phytoliths was achieved by add-
ing 5 ml of sodium polytungstate solution at 2.4 g/ml density.
Phytoliths were mounted on microscope slides using glycerol
and observed under ×400 magnification using a Zeiss CP-
achromat light microscope mounted with a camera to identify
the taxonomically important phytolith morphotypes.
Phytolith morphotypes were classified following pub-
lished micrographs and descriptions by various authors and
wherever possible named according to the ICPN Working
Group of Madella et al. (2005), and interpretations of their taxo-
nomic origin are also based on published data. Poaceae
phytolith morphotypes (grass silica short cells – GSSCs) were
classified according to Twiss et al. (1969), Alexandre et al. (1997),
Piperno (2006), Barboni and Bremond (2009), while phytoliths
produced by non-Poaceae were classified according to Runge
(1999), Albert et al. (1999), Piperno (2006); Mercader et al. (2009),
amongothers. Phytolith abundances were gauged byscanning
one vertical column of each slide, and the phytolith morpho-
type count ranged from 360 to 680 in the samples analysed.
Because of issues of multiplicity (the production of several
different phytolith forms within a single plant species) and
redundancy (similar phytolith forms being produced by
several plant species) (Rovner 1971), it is difficult to assign a
definite taxonomic classification to a specific morphotype.
However, some morphotypes may be attributed to particular
taxa, for instance grass subfamilies can be identified by their
quantities, size, their occurrence in combination with other
morphotypes, and depending on their environmental settings
(Twiss 1992; Piperno 2006; Barboni & Bremond 2009; Table 2).
Phytolith types were grouped into ten main categories
(Table 2). Of these, the presence of four morphotypes, which
areproduced exclusively in epidermal short cells of grasses and
are of known taxonomic significance (Twiss et al. 1969; Piperno
2006), allows for the identification of some of the main grass
subfamilies: (i) cross; (ii) bilobate (the few cross-types present
aregrouped with bilobates in this study); (iii) saddle type (short
and long); and (iv) rondel-shaped. The cross and bilobate
morphotypes predominate in the subfamily Panicoideae; the
saddles are dominant in the subfamily Chloridoideae and
rondels are typically associated with subfamily Pooideae
(Table 2). Phytoliths that occur in all members of the grass
familyand other monocots are: (v) cuneiform bulliform; as well
as (vi) elongate types (psilate and echinate long cells; tabular
andcylindrical elongates) that are also produced by some trees,
or dicots in general (Piperno 1988; Thorn 2001; Mercader et al.
2009; Novello, Barboni, Berti-Equille, Mazur, Poilecot &
Vignaud 2012). Other non-Poaceae phytoliths were grouped
as: (vii) globular granulate that is often associated with trees/
dicots (Alexandre et al. 1997; Runge 1999; Mercader et al. 2009);
(viii) the parallelepiped blocky types that are usually associated
with wood/bark of woody species (Albert et al. 1999; Mercader
et al. 2009); (ix) the hat/cone-shaped and achene phytolith
morphotypes that are characteristic of Cyperaceae (sedges)
(Piperno 1988, 1989; Ollendorf 1992); and (x0) the globular
psilate that have several origins occurring in both monocots
and dicots and are not useful in taxonomic discrimination
(Piperno 1988, 2006). Table 2 shows the nomenclature of the
main phytolith categories observed and their plant attribution
according to the literature.
Phytoliths are well preserved and are abundant in all four
samples analysed. The micrographs of some of the phytolith
morphotypes are shown in Fig. 5. The relative abundance
(%) of each morphotype is represented in Fig. 6. The counts
presented here were reached by counting one vertical column
only of the microscope slides. Poaceae morphotypes are partic-
ularly useful in separating some of the grass subfamilies and
their habitats. Here they represent more than 40% in all
134 South African Archaeological Bulletin 71 (204): 130–145, 2016
South African Archaeological Bulletin 71 (204): 130–145, 2016 135
samples apart from sample N4 SE Layer 2 (top) with 37%, with
the highest contributor in all four samples being the rondel-
shaped morphotypes (27–35%, Fig. 5E,F). This type was, in the
past, generally attributed to C3high altitude/cold climate
grasses of the subfamily Pooideae (Twiss et al. 1969; Twiss
1992); however, rondels are now considered to be the most
redundant phytolith morphotype occurring in all the main
grass subfamilies (Bamford et al. 2006; Barboni & Bremond
2009; Mercader et al. 2010; Novello et al. 2012; Cordova 2013).
Rondels were found to occur in large numbers (up to 95%) in
the genera Sporobolus and Eragrostis (subfamily Chloridoideae,
hot dry climate grasses) by Novello et al. (2012) in agreement
with the previous aforementioned studies in Africa. Because
the trapeziform sinuate morphotype that is reported to con-
firm the presence of C3Pooideae cold climate grasses was not
observed in the samples (Barboni & Bremond 2009), these
morphotypes are most likely to have been produced by C4
Bilobate-shaped (mainly C4tall Poaceae of warm wet
climates – subfamily Panicoideae, Fig. 5B) and saddle-shaped
(C4short Poaceae of dry climates – subfamily Chloridoideae,
Fig.5C,D) phytoliths occur in almost similar amounts with each
only contributing 4–6% of the total count with the saddles
reaching 10% in one sample. Globular granulate (Fig. 5H)
morphotypes that are typically associated with trees/dicots
seem to be the second most abundant phytolith type, varying
from 23 to 39% and they are used to provide information on
the relative density of woody plants within the vegetation
type (Alexandre et al. 1997). Cyperaceae achene phytoliths
(Fig. 5L,M,Q), which appear similar to the Scirpus-type and
unknown possible Cyperus-type achenes described by Piperno
(1989) and Iriarte et al. (2010) were observed in low frequencies.
Phytolith achenes typical of the genus Cyperus and Kyllinga
(Piperno 1989; Iriarte et al. 2010) were observed, but were
very rare compared to the ‘Scirpus’ type. Hat-shaped/cone
phytoliths that are typical of Cyperaceae (sedges) together
with these achene phytoliths are represented in low frequen-
cies (1 to 2%). Another type of globular psilate morphotype was
encountered (Fig. 5O); its surface appeared either rough or
smooth with a central depression.
Parallelepiped blocky morphotypes (Fig. 5J) occur in per-
centages that range between about 8% and 17%. They are often
associated with the bark/wood of woody species, but they have
been reported to occur in some sedges and grasses (Novello
et al. 2012). They were observed in some woody species of
Fabaceae, Ebenaceae and Clusiaceae in Mozambique (Mercader
et al. 2009). In addition, elongate tabular morphotypes with a
‘laminate median swelling’ (Fig. 5R) similar to those observed
by Novello et al. (2012), which are associated with herbaceous
plants, were observed in low frequencies, and it is not clear if
Fig. 5P is a broken Fig. 5R morphotype or just a short variant of
the same. One globular decorated morphotype (Fig. 5N) that is
similar to globular echinates typical of palms (Piperno 2006)
was observed. Only side protrusions are clear in the morpho-
THE LITHIC ANALYSIS
Most of the lithic pieces in the three layers studied are
patinated so we developed gradations of patination (see next
section), and we recorded the percentage of surface that is
patinated (Table 3). Greater patination occurs in the oldest
layer, thus, as can be observed in Table3, about 80% of the Layer
3 lithics have heavy Degree 3 patina. The differences in patina-
tion through time suggest that the three layers have some
stratigraphic integrity, notwithstanding the evidence for
weathering and trampling. This interpretation is supported by
theuneven densities of lithics through time. The base of Layer 1
and the base of Layer 2 have higher lithic densities than the
TABLE 2.Main phytolith morphotypes identified in the Steenbokfontein sequence with their taxonomic attributions according to the literature.
Morphotypes Main taxonomic attribution Reference
Cross and bilobate Poaceae – Panicoideae, Twiss et al.1969;
Arundinoideae/Danthonioideae Fredlund & Tieszen, 1994;
Mercader et al. 2010
Saddle Poaceae – Chloridoidae, Twiss et al. 1969;
Arundinoideae/Danthonioideae Mercader et al. 2010
Rondel Poaceae – Pooideae, Chloridoideae Twiss 1992;
Bamford et al. 2006;
Barboni & Bremond 2009;
Novello et al. 2012;
Cuneiform bulliform Poaceae Twiss 1992
Elongate (psilate, echinate, tabular, Monocots, Dicots Piperno 1988;
cylindrical) Thorn 2001;
Mercader et al. 2009;
Novello et al. 2012
Globular granulate Dicots Alexandre et al. 1997;
Parallelepiped blocky Dicots, some Monocots Albert et al. 1999;
Mercader et al. 2009;
Novello et al. 2012
Hat/cone-shaped and achene Cyperaceae Piperno 1989;
phytoliths Ollendorf 1992
Globular psilate Monocots, Dicots Piperno 2006
middle or top of either of these layers, and Layer 3 has the
lowest lithic density of any of the layers (Table 1). If the three
layers were the result of a dumped deposit, we should have
expected patination to be mixed through all layers and for the
densities of lithics to be similar.
A preliminary analysis of the Steenbokfontein assemblage
highlighted four important characteristics which influenced
the subsequent analysis and interpretation. First, it is made
almost entirely from silicified siltstone which outcrops at the
spring. Secondly, some flakes are geological, not anthropogenic.
Raw blocks of siltstone at the site are rectangular slabs weath-
ered from the highly jointed outcrop. Many blocks are missing
corners that appear to have been removed by mechanical
processes. These natural removals result in pseudo-flakes and
pseudo-cores. Any piece without a clear platform and bulb of
percussion was eliminated from the sample analysed. Thirdly,
most of the silicified siltstone slabs and flakes display various
degrees of white patina, probably produced by alternate sub-
mersionin water when the spring level was high, and exposure
to sun and wind when water levels were low. Some of the
pieces have white patina only on one face, suggesting partial
burial and exposure. Finally, regardless of their morphology
or the extent of weathering, most pieces appear damaged.
Indeed, all the stone tools have secondary edge modification
likely to have been produced by human and animal trampling.
Trampling characteristics are: (i) frequent crushing (particu-
larly in Layer 1) (Table 3); (ii) dull or blunt edges on half the
pieces (Table 3); (iii) extensive secondary edge modification on
pieces from Layers 1 and 2; (iv) 60% of all lithics have more than
three directions of secondary edge modification and, (v) a high
percentage of irregular, dispersed scars on blank edges. In
order to test this hypothesis we performed an experimental
programme to compare truly retouched lithic assemblages to
trampled assemblages and we later compared this to the
136 South African Archaeological Bulletin 71 (204): 130–145, 2016
FIG. 5.Microphotographs of phytoliths from the Steenbokfontein sediment samples analysed. Scale bars are 10 mm except for C, F, K,R=5mmandQ=20mm.
(A)cross; (B)Bilobate; (C)short saddle; (D)long saddle; (E)rondel; (F)rondel; (G)cuneiform bulliform; (H)globular granulate; (I)elongate – cylindrical;
(J)parallelepiped – blocky; (K)elongate – tabular; (L) Cyperaceae achene;(M) Cyperaceae achene;(N)globular decorated – echinate type? (O)globular
psilate with central depression; (P)tabular knobbed; (Q)articulated Cyperaceae achene;(R)elongate tabular knobbed.
South African Archaeological Bulletin 71 (204): 130–145, 2016 137
archaeological material of this site. This experimental programme
is research in progress.
THE LITHIC TECHNOLOGICAL ATTRIBUTE ANALYSIS
We employ the chaîne opératoire approach (Karlin, Bodu &
Pelegrin 1991), which views an assemblage as the outcome of
cultural choices. As several scholars have pointed out (for
example Dibble 1995; Shott 2003; Tostevin 2013), the chaîne
opératoire approach can be highly subjective if a purely qualita-
tive, systemic approach is used. In order to make this study as
objective as possible, we incorporate various quantitative
parameters. Furthermore, after the attribute analysis, we use
basic statistics to support the qualitative arguments.
The large lithic collection from the Surface layer was
excluded because we assumed these lithics were disturbed,
and we analysed lithics only from Layers 1–3. We established
the cut-off between chips and blanks as 2 cm. The chips include
pieces with a wide range of morphologies and they are not
studied further here. For the 273 pieces >2cm, we analysed
both complete and fragmented pieces.
The lithic assemblage was divided into three broad analytical
categories: (i) cores; (ii) blanks with or without secondary edge
modification; and (iii) chips. Variables recorded for the techno-
logical analysis are described in de la Peña (2015: table 3).
In order to assess the taphonomic alteration of the lithics,
we used the following macro trace parameters:
•Presence/absence of crushing/fissuration: Crushing is defined
as steep step and hinge fracture resulting from contact with
the edge of the flake at an angle approaching 90 degrees.
•Degree of bluntness of the edges (sharp, dull and blunted):
The whole edge of the piece is assessed for relative bluntness.
If it is fresh it is called sharp. If the piece has lost some of its
sharpness, it is called dull. If it has rounded edges and/or
steep sides from crushing/trampling, it is called blunt.
•Position of secondary edge modification relative to a particular
edge. The maximum number of positions is ten because we
count secondary edge modification that appears on the
central ridge of the dorsal face of flakes.
• Regularity (coherent/incoherent pattern of scars): Regularity
refersto the overall similarity (size,shape, depth) and general
resemblance of different secondary edge modification scars
to each other on one piece.
• Continuity (isolated, dispersed, continuous): This refers to
the location and grouping of secondary edge modification.
‘Isolated’ refers to a single secondary edge modification scar
whereas ‘continuous’ secondary edge modification is sequen-
tial along an edge.
•Presence/absence of edge fracture: This records broken flake
edges.A simple exampleis the distal tip of a convergent flake.
•The predominant morphology of scars (semicircular/half-
moon, quadrangular, trapezoidal, triangular, irregular)
(González-Urquijo & Ibáñez Estévez 1994).
•The types of scar morphology: this parameter scores the
number of scar morphology types.
•Degree of patina: 0 = no patina, 1 = slightly patinated, 2 =
medium patination, 3 = entirely patinated.
The three layers share technological characteristics so they
are described together. Later we use metrics to investigate
potential distinctions. Three of the cores represented in layer 1
(Fig. 7A,B,C) are prismatic blade cores. The siltstone naturally
appears in rectangular blocks, and this type of natural
morphology was ideal for blade prismatic reduction. The
fourth core is a core on flake for bladelets (Fig. 7D).
Blade prismatic production is well represented in the
assemblage, not only by the prismatic cores already mentioned,
but by many blade blanks with rectilinear profiles, trapezoidal
cross-sections and predominantly unidirectional scar patterns
(Figs 8 & 9). Most of these blanks were obtained without
platform preparation because most of the platforms are plain
or cortical (Fig. 9). Faceted platforms may be ‘pseudo’ ones, as
FIG. 6.Histograms showing Steenbokfontein phytolith abundance in sediment samples from Square N4.
they were probably produced through an a posteriori trampling
process; originally, they might have been plain platforms.
Bladeletproduction is only represented by the core on flake
(Fig. 7D), although it is not so clearly represented among the
blanks, as only five small fragmented pieces among the chip
assemblagewere identified as bladelets(from Layers 1 and 2).
The second knapping method is a centripetal reduction of
the rectangular blocks in order to produce flakes. The blanks
coming from this type of knapping method have crossed,
centripetal or sub-centripetal scar patterns, and elongated
triangular and right triangular cross-sections. Again, it seems
that this type of knapping method was accomplished without
preparation of the platforms, because plain and cortical plat-
forms are common in the assemblage whereas dihedral
platforms have low representation (Fig. 9). The flake knapping
method seems quite opportunistic and appears to have relied
on the morphology of the blocks. Therefore, it does not seem
that the knapping method can be viewed as a purely discoidal
one. Moreover, there are no typical discoidal by-products like
pseudo-Levallois points or side (core) flakes showing a 45
degree angle between dorsal face and platform (typical of
discoidal reduction) (Boëda et al. 1990). The types of flake
blanks represented seem to fit more into a multi-facial core
definition without an established pattern. Most of the blanks
with centripetal and sub-centripetal scar patterns and expand-
ing shapes should belong to this unstandardised knapping
method (Figs 8 & 9). In contrast, most of the unidirectional
blanks and unidirectional convergent blanks with parallel
edges should belong to blade prismatic reduction and, as can
be seen in Fig. 9, this includes more than half of the blanks.
Moreover, part of the flake production should belong to the
preparation and trimming of the blade prismatic cores. It
should be highlighted that one point might be tentatively
assigned to a Levallois reduction framework (Fig. 8), but no
other blanks fit this knapping method.
Another general insight into the lithic assemblage is the
small number of dorsal scars on the blanks (Fig. 9), which could
mean that the knapping activity at the site might have been
mainly for testing blocks, and for preliminary phases of
knapping. This could mean that Steenbokfontein was a knapping
workshop. The high representation of cortical platforms and
cortical flakes also supports this hypothesis (Fig. 9).
Regarding formal tools, it is very difficult to give an idea of
how many there may once have been because it is impossible to
discriminate between purposeful retouch and secondary edge
modification caused by post-depositional trampling. Some
examples of equivocal formal tools are shown in Fig. 8 together
with the point.
Table 4 demonstrates that flakes are predominant in the
lithic assemblage. One of the questions that can be answered
quantitatively is whether the typometrical distribution of
blades and flakes is normal or not. In the event that it is normal,
this could mean that, for each of the knapping methods, there
was continuous reduction. Whereas if not normal, this might
imply several knapping methods, or that the types of blanks
produced had very different typometric sizes.
In Fig. 10 the length, breadth and thickness distributions
for Layer 1 and 2 flakes and blades are shown (Layer 3 is
excluded because it has few items). For flakes, the size distribu-
tionis very similar in the two layers, whereas for blades, Layer 1
seems to have larger blades than Layer 2, and possibly a
As can be seen in Table 5, from the Shapiro-Wilk normality
tests, and the attributes of length and breadth, the distributions
for Layer 1 flakes and blades are not normal. On the contrary,
Layer 2 has normal distributions. This situation seems to rein-
force the idea, already highlighted from the Layer 1 blade
histograms (Fig. 10), that at least for the blades we might have
two blank sizes. Thus we proceed to a mixture analysis in order
to assess this possibility. Mixture analysis is an agglomerative,
hierarchical and univariate statistical test. It is a method of
maximum likelihood estimation to recognise parameters of two
138 South African Archaeological Bulletin 71 (204): 130–145, 2016
TABLE3.Attributes of Steenbokfontein lithics from L ayers 1 to 3. Percentages
of degree of patina, percentage of patina, crushing, degree of bluntness, position
of secondary edge modification, regularity, continuity, scars in fracture and
Layer 1 Layer 2 Layer 3
Degree of patina (%)
0 12.15 8.51 0.00
1 45.79 0.00 5.56
2 32.71 44.68 16.67
3 9.35 46.81 77.78
Percentage of patina
0 11.32 8.51 0.00
1 to 10 0.94 0.00 0.00
11 to 40 7.55 4.26 11.11
41 to 60 37.74 21.28 0.00
61 to 90 16.98 27.66 0.00
91 to 99 17.92 0.00 0.00
100 7.55 38.30 88.89
Yes 50.26 17.31 34.62
No 49.74 82.69 65.38
Degree of bluntness (%)
Blunt 3.59 9.62 3.85
Dull 54.87 38.46 46.15
Sharp 41.54 51.92 50.00
Position of secondary edge
1 3.23 6 10
2 12.37 14 10
3 13.98 16 20
>3 70.43 64 60
Irregular 83.98 90 100
Regular 16.02 10 0
Continuous 38.80 10 5
Dispersed 56.28 84 70
Isolated 4.92 6 25
Scars in fracture (%)
Yes 9.74 32.69 0
No 90.26 67.31 100
Edge fracture (%)
Yes 19.57 34.62 47.83
No 80.43 65.38 52.17
TABLE 4.Technological categories at Steenbokfontein by layer.
Chips are excluded from the percentages.
Cores Flakes Blades Chips
n(%) n(%) n
Layer 1 4 125 (65.4) 66 (34.6) 482
Layer 2 0 29 (55.7) 23 (44.3) 85
Layer 3 0 10 (38.5) 16 (61.5) 115
Total n4 164 105 682
South African Archaeological Bulletin 71 (204): 130–145, 2016 139
FIG. 7. Cores from Layer 1, Steenbokfontein. (A,B,C)Prismatic blade cores; (D)core on flake to produce bladelets. Scale: bars =1 cm.
or more univariate normal distributions grouped in a single
sample (Barceló 2007). In this case, we applied the mixture
analysis to the breadth readings for flakes and blades of Layer
1. In both cases, two groups were the most likely scenario,
rather than one or three groups. In Table 6 we show the proba-
bilities and the standard deviations of the groups.
Relatively few MSA sites have been studied on the Waterberg
plateau and none is dated. Steenbokfontein is the first MSA
open site excavated on the plateau. In contrast, several late LSA
sites have been excavated here (Van der Ryst 1998). Unfortu-
nately, Steenbokfontein could not be dated because sand
grains of mixed age were present.
MSA stone tools were the only archaeological finds in the
four layers excavated at Steenbokfontein. Petrographic analysis
and XRF concur that the rock used was silicified siltstone. All
the stone tools seem to have been made from the rock outcrop
at the spring and they appear mixed with desegregated
siltstone regolith as though they were knapped on top of the
regolith and then exposed and trampled for years or even
millennia. Mechanical secondary edge modification occurs on
both archaeological and geological material and this presented
major obstacles to the analysis of the stone tools because it was
difficult to distinguish confidently between knapping, espe-
cially retouch, and accidental conchoidal fractures. In order to
test our trampling hypothesis, we have designed a trampling
experiment with geological siltstone from Steenbokfontein
and this is the subject of a separate study.
At Steenbokfontein we have documented two main
140 South African Archaeological Bulletin 71 (204): 130–145, 2016
FIG. 8. Steenbokfontein blades, flakes, and lithics with secondary edge modification. Top left: blade blanks from Layer 1. Top right: flake blanks from Layers 1
and 2; Bottom: Examples of Steenbokfontein lithics with secondary edge modification. (A)and (C)are blade proximal fragments with secondary edge modification
in their distal fractures; (D)point.
South African Archaeological Bulletin 71 (204): 130–145, 2016 141
knapping methods: prismatic blade production and centripetal
flake production. In Layer 1 it seems that there were different
typometric size objectives for flakes and blades, whereas in
Layer 2 we detected a normal distribution for flakes and blades,
which probably means a continuous reduction process. The
high proportion of cortex and the small number of scars on
flake dorsal faces might be pointing towards a workshop,
perhaps for testing rock slabs. We hypothesise, further, that
this site was probably a knapping workshop where only initial
knapping and testing of geological blocks took place. This
makes Steenbokfontein unusual among the Limpopo sites
discussed here. Not only is it an open site with lithics made on
rocks that do not occur elsewhere in Limpopo, but it appears to
be a special purpose destination. As explained at the beginning
of the paper, the silicified siltstone used here has extremely
limited geological occurrence. Notwithstanding this, LW has
observed MSA tools made on silicified siltstone more than
20 km from the nearest outcrop, so rocks and/or lithic products
were transported to other sites on the Waterberg plateau. It
would be worth conducting a field survey to plot the distribu-
FIG. 9.Attributes of blanks from Steenbokfontein.(A)Percentage of scar pattern, shape of the blank, cross section and type of platforms in Layers 1–3; (B)shape of
the blank and scar pattern in Layers 1 and 2; (C)percentage of cortex and number of scars on the dorsal face of lithics in Layers 1 and 2.
142 South African Archaeological Bulletin 71 (204): 130–145, 2016
FIG. 10.Histograms of length, breadth and thickness of Steenbokfontein flakes and blades from Layers 1 and 2.
South African Archaeological Bulletin 71 (204): 130–145, 2016 143
tions of these relatively rare rocks. Visits to sites like Steenbok-
fontein were probably part of the regular cycle of hunter-
gatherers that lived in the area. Steenbokfontein thus adds to
our understanding of variable site use in the MSA.
Our description of the Steenbokfontein lithic assemblage
makes it plain that the site cannot readily be placed within the
Pietersburg Industry, not least because, apart from Cave of
Hearths, descriptions of Pietersburg assemblages are too
general to be useful. Pietersburg is epitomised by large elon-
gated products, including long points that are usually unifacial
and manufactured on blades (Mason 1962; Sampson 1974).
Cores and end products are often made on hornfels (Mason
1962; Sampson 1974), a rock that sometimes occurs in large
blocks that allow the knapping of long products. Other rocks
that occur in large pieces, such as quartzite, were also used,
suggesting that the variable appearance of Pietersburg assem-
blages may, to a degree, be influenced by available rocks.
Although the Steenbokfontein assemblage does contain many
blades, they are not as large as the ones appearing in the
Pietersburg sites described by Mason and Sampson. The use of
silicified siltstone is not recorded in any Pietersburg assemblage,
so it is possible that the morphology of Steenbokfontein’s
assemblage has been constrained by this rock. Furthermore, if
the site is a primary workshop, large products may have been
transported to other sites. Our small excavation yielding only
four cores cannot resolve the issue; further excavation and a
larger lithic sample is required for this.
All four layers at Steenbokfontein contain similar phytolith
morphotypes representing C4 grasses, some Cyperaceae, and
woody plants. Since grasses produce huge numbers of
phytoliths they tend to be over represented (Piperno 2006) so
the vegetation characterised is probably more wooded than
grassy, especially Level 2 (top) where there is a high proportion
ofglobular granulates that are produced bydicots (Runge 1999;
Mercader et al. 2009). No Cyperaceae grow around the spring
now, so wetter conditions may have prevailed. Since there are
no dates for Steenbokfontein, its environmental record cannot
be compared meaningfully with other records. Nevertheless,
its record is not unlike that from the spring site Wonderkrater.
Here charcoal, phytolith and pollen data imply that MIS6 occu-
pations took place in warm, dry grassy savanna woodland and
that the MIS3 conditions were cooler and wetter with extensive
grassland with some woody shrubs (Backwell et al. 2014).
Other botanical remains at Steenbokfontein were restricted
to burnt grass roots that intruded into the artefact-bearing
layers. These illustrate the important taphonomic issue that
seemingly in situ carbonised remains are not necessarily con-
temporary with the layer in which they occur. Experiments
with fire have demonstrated heat transfer to substrates 10 cm
below a fire; these carbonise buried organic material according
to the temperature of the surface fire and its duration (Sievers
& Wadley 2008; Aldeiasa et al. 2016). No hearth features were
observed at Steenbokfontein so the grass roots recovered
through flotation are likely to have been carbonised by surface
grass fires, none of which has come through the area in the past
The small excavation at Steenbokfontein 9KR was con-
ducted close to a spring eye, one of several on the farm of the
same name. The uppermost stratum of silt was immediately
underlain by siltstone rubble embedded in silt. MSA lithics
made on silicified siltstone were recovered throughout the
sequence, from the surface to bedrock. Far fewer lithics were
recoveredfrom the top of bedrock (Layer 3) than fromthe other
layers. They were often recovered lying vertically or at angles
other than horizontal. This implies disturbance of the sedi-
ments, an interpretation supported by the mixture of sand
grains that precluded OSL dating, and by the damage to lithics
and unworked rock fragments. The secondary edge modifica-
tion on archaeological lithics and siltstone pieces of geological
origin is consistent with trampling by humans and animals
when visiting the spring. Mechanical modification is likely to
have taken place from the MSA to recent times and it has obvi-
ously influenced the distribution, not only of lithics, but also
sand grains and phytoliths. Furthermore, the desegregated
siltstone fragments and the stone tools have been affected by
water, presumably as a result of fluctuating water levels in the
spring. There is a water patina on most of the archaeological
and geological fragments. Therefore, the major factors that
altered archaeological material from Steenbokfontein are tram-
pling, the change in the level of the water from the spring, and
the long exposure of the material to the elements. Changing
water levels may also have removed some original sediment
from the rock-strewn layers.
Our preliminary excavation at the Steenbokfontein spring
suggests that further work would be profitable in the vicinity.
The high abundance and excellent preservation of phytoliths
implies that further phytolith research could produce interest-
ing vegetation studies at this site and its surrounds. Targeting
an area away from the spring might also produce a less
damaged lithic assemblage, and a sediment column with integ-
rity that can be OSL dated.
Steenbokfontein cannot readily be compared to other MSA
sites in Limpopo. This is not only because of the small sample of
lithics analysed here, the post-depositional damage to the
TABLE 6.Mixture analyses of breadths of flakes and blades in Layer 1,
Breadth flakes (two groups). Log l.hood: –297.4; Akaike IC: 603.1
Probability Mean S.D.
0.62704 23.086 4.4579
0.37296 33.905 8.0467
Breadth blades (two groups) Log l.hood: –156.4; Akaike IC: 321.4
Probability Mean S.D.
0.43051 28.022 7.3363
0.56949 16.934 3.6422
TABLE 5.Shapiro-Wilk normality tests conducted on lengths and breadths of flakes in Layers 1 and 2, Steenbokfontein. * = cases that are not normal.
Length flakes Breadth flakes Length flakes Breadth flakes Length blades Breadth blades Length blades Breadth blades
Layer 1 Layer 1 Layer 2 Layer 2 Layer 1 Layer 1 Layer 2 Layer 2
n82 119 17 28 24 64 9 23
Shapiro-Wilk (W) 0.9361 0.9469 0.9774 0.9378 0.8891 0.9279 0.9439 0.9466
P(normal) 0.0005047* 0.0001382* 0.9293 0.09725 0.01275* 0.001084* 0.6234 0.2479
144 South African Archaeological Bulletin 71 (204): 130–145, 2016
lithics, and the fact that silicified siltstone was not available at
sites below the Waterberg plateau. Steenbokfontein is almost
certainly a lithic workshop where blocks of rock were tested
and knapped in a preliminary way. Recent work in southern
Africa has revealed considerable variation in MSA sites across
space as well as through time. Steenbokfontein adds to that
variability and demonstrates the importance of studying open
sites as well as the rich cultural traps created in caves and rock
We thank the Van der Walt family for allowing excavation
on Steenbokfontein 9KR, and for their kind hospitality. The site
was excavated with SAHRA permit #1891. After completion of
the project, the material will be housed in the Polokwane
Museum, Limpopo. We are grateful to Zenobia Jacobs for
attempting to date the site. L.W. receives funding from the
National Research Foundation, but opinions expressed here
are not necessarily those of the funding agency. The University
of the Witwatersrand and the Evolutionary Studies Institute
are thanked for their ongoing support of archaeological
research. P.d.l.P has a post-doctoral fellowship supported by
the Centre of Excellence in Paleosciences. We are grateful to
Lucinda Backwell, Maria van der Ryst, and an anonymous
reviewer for useful comments that improved this paper.
Albert, R.M., Tsatskin, A., Ronen, A., Lavi, O., Estroff, L., Lev-Yadun, S.
& Weiner, S. 1999. Mode of occupation of Tabun Cave, Mt Carmel,
Israel, during the Mousterian Period: a study of the sediments and
phytoliths. Journal of Archaeological Science 26: 1249–1260.
Aldeiasa, V., Dibble, H.L., Sandgathe, D. & Goldberg, P. 2016. How heat
alters underlying deposits and implications for archaeological fire
features: a controlled experiment. Journal of Archaeological Science 67:
Alexandre, A., Meunier, J.D., Lezine, A.M., Vincens, A. & Schwartz, D.
1997. Phytoliths: indicators of grassland dynamics during the late
Holocene in intertropical Africa. Palaeogeography, Palaeoclimatology,
Palaeoecology 136: 213–229.
Backwell, L., McCarthy, T.S., Wadley, L., Henderson, Z., Steininger,
C.M., de Klerk, B., Barré, M., Lamothe, M., Chase, B.M., Woodborne,
S., Susino, G.J., Bamford, M.K., Sievers, C., Brink, J.S., Rossouw, L.,
Pollarolo, L., Trower, G., Scott, L. & d’Errico, F. 2014. Multiproxy
record of late Quaternary climate change and Middle Stone Age
human occupation at Wonderkrater, South Africa. Quaternary Science
Reviews 99: 42–59.
Bamford, M.K., Albert, R.M. & Cabanes, D. 2006. Plio-Pleistocene
macroplant fossil remains and phytoliths from Lowermost Bed II in
the eastern palaeolake margin of Olduvai Gorge, Tanzania. Quater-
nary International 148: 95–112.
Barboni, D. & Bremond, L. 2009. Phytoliths of East African grasses: an
assessment of their environmental and taxonomic significance based
on floristic data. Review of Palaeobotany and Palynology 158: 29–41.
Barceló, J.A. 2007. Arqueología y estadística: Introducción al estudio de la
variabilidad de las evidencias arqueológicas. Barcelona: Universitat
Autònoma de Barcelona. Servei de Publicacions.
Barré, M., Lamothe, M., Backwell, L. & McCarthy, T. 2012. Optical
dating of quartz and feldspars: a comparative study from Wonder-
krater, a Middle Stone Age site of South Africa.Quaternary Geochronol-
ogy 10: 374–379.
Boëda, E., Geneste, J-M. & Meignen, L. 1990. Identification des Chaînes
opératoires lithiques du Paléolithique Ancien et Moyen. Paléo 2:
Callaghan, C.C. 1993. The geology of the Waterberg group in the south-
ern portion of the Waterberg Basin. Geological Survey Bulletin 104:
Cordova, C.E. 2013. C3Poaceae and Restionaceae as potential proxies
for reconstructing winter rainfall in South Africa. Quaternary Interna-
tional 287: 121–140.
De Bruiyn, H. 1971. The geology of the central portion of the Waterberg
basin around Vaalwater, Northern Transvaal. Unpublished report of
the Geological Survey of South Africa, 1971-0044: 1–11.
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 Vries, W.C.P. 1970. Stratigraphy of the Waterberg System in the
southern Waterberg area, northwestern Transvaal. Annals of the Geo-
logical Survey of South Africa 7: 43–67.
De Vries, W.C.P. 1973. Sedimentary structures in the southern and cen-
tral portions of the Waterberg area, northwestern Transvaal. Geologie
en Mijnbouw 52: 133–140.
Dibble, H.L. 1995. Biache Saint-Vaast, Level IIA: a comparison of analyt-
ical approaches. In: Dibble, H. & Bar Yosef, O. (eds) The Interpretation
and Definition of Levallois Variability: 93–116. Madison: Prehistory
Fredlund, G. & Tieszen, L.T. 1994. Modern phytolith assemblages from
theNorth American Great Plains. Journal of Biogeography 21: 321–335.
González-Urquijo, J.E. & Ibáñez Estévez, J. 1994. Metodología de análisis
funcional de instrumentos tallados en sílex. Bilbao: Universidad de
Hutson, J.M. & Cain, C.R. 2008. Reanalysis and reinterpetation of the
Kalkbank faunal accumulation, Limpopo Province, South Africa.
Journal of Taphonomy 6: 399–428.
Jansen, H. 1982. The geology of the Waterberg Basins in the Transvaal,
Republic of South Africa. Geological Survey Memoir 71: 1–98.
Karlin, C., Bodu, P. & Pelegrin, J. 1991. Processus techniques et chaînes
opératoires. Comment les préhistoriens s’appropient un concept
élaboré par les ethnologues. In: Balfet, H. (ed.) Observer l’action tech-
nique, les chaînes opératoires, pour quoi faire?:101–117. Paris: CNRS.
Iriarte, J., Glaser, B., Watling, J., Wainwright, A., Birk, J.J., Renard, D.,
Rostain, S. & McKey, D. 2010. Late Holocene neotropical agricultural
landscapes: phytolith and stable carbon isotope analysis of raised
fields from French Guianan coastal savannahs. Journal of Archaeologi-
cal Science 37: 2984–2994.
Madella, M., Alexandre, A. & Ball, T. 2005. ICPN Working Group. Inter-
national Code for Phytolith Nomenclature 1.0. Annals of Botany 96:
Mason, R.J. 1962. Prehistory of the Transvaal. Johannesburg: Witwaters-
rand University Press.
Mason, R.J. 1988. Cave of Hearths, Makapansgat, Transvaal. Archaeologi-
cal Research Unit Occasional Paper 21. Johannesburg: University of
Mercader, J., Astudillo, F., Barkworth, M., Bennett, T., Esselmont, C.,
Kinyanjui, R., Laskin Grossman, D., Simpson, S. & Walde, D. 2010.
Poaceae phytoliths from the Niassa Rift, Mozambique. Journal of
Archaeological Science 37: 1953–1967.
Mercader, J., Bennett, T., Esselmont, C., Simpson, S. & Walde, D. 2009.
Phytoliths in woody plants from the Miombo woodlands of Mozam-
bique. Annals of Botany 104: 91–113.
Novello, A., Barboni, D., Berti-Equille, L., Mazur, J., Poilecot, P. &
Vignaud, P. 2012. Phytolith signal of aquatic plants and soils in Chad,
Central Africa. Review of Palaeobotany and Palynology 178: 43–58.
Ollendorf, A.L. 1992. Toward a classification scheme of sedge
(Cyperaceae) phytoliths. In: Rapp Jr., G. & Mulholland, S.C. (eds)
Phytolith Systematics: Emerging Issues: 91–111. New York: Plenum
Pearsall, D.M. 2000. Paleoethnobotany: A Handbook of Procedures, 2nd edn.
New York: Academic Press.
Piperno, D.R. 1988. Phytolith Analysis: An Archaeological and Geological
Perspective. California: Academic Press.
Piperno, D.R. 1989.The occurrence of phytoliths in the reproductive
structures of selected tropical Angiosperms and their significance in
tropical paleoecology, paleoethnobotany and systematics. Review of
Palaeobotany and Palynology 61: 147–173.
Piperno, D.R. 2006. Phytoliths. A Comprehensive Guide for Archaeologists
and Paleoecologists. Lanham, New York, Toronto, Oxford: Altamira
Plug, I. 1981. Some research results on the late Pleistocene and early
Holocene deposits of Bushman Rock Shelter, eastern Transvaal.
South African Archaeological Bulletin 36: 14–21.
Porraz, G., Val, A., Dayet, L., de la Peña, P., Douze, K., Miller, C.E.,
Murungi, M.L., Tribolo, C., Schmid, V.C. & Sievers, C. 2015. Bushman
Rock Shelter (Limpopo, South Africa): a perspective from the edge of
the Highveld. South African Archaeological Bulletin 70: 166–179.
Rovner, I. 1971. Potential of opal phytoliths for use in paleoecological
reconstruction. Quaternary Research 1: 343–359.
South African Archaeological Bulletin 71 (204): 130–145, 2016 145
Runge, F. 1999. The opal phytolith inventory of soils in central Africa –
quantities, shapes, classification, and spectra. Review of Palaeobotany
and Palynology 107: 23–53.
Sampson, G.S. 1974. The Stone Age Archaeology of Southern Africa. New
York and London: Academic Press.
Schoonraad, M. & Beaumont, P.B. 1968. The North Brabant Shelter,
north western Transvaal. South African Journal of Science 64: 319–331.
Shott, M.J. 2003. Chaîne opératoire and reduction sequence. Lithic Tech-
nology 28: 95–105.
Sievers, C. & Wadley, L. 2008. Going underground: experimental
carbonization of fruiting structures under hearths. Journal of Archaeo-
logical Science 35: 2909–2917.
Sinclair, A. 2009. The MSA stone tool assemblage from the Cave of
Hearths, Beds 4–9. In: McNabb, J. & Sinclair, A. (eds) The Cave of
Hearths: Makapan Middle Pleistocene Research Project: Field Research by
Anthony Sinclair and Patrick Quinney, 1996–2001: 105–137. Oxford:
Sumner, T.A. 2013. A refitting study of late Early to Middle Stone Age
lithic assemblages from the site of Kudu Koppie, Limpopo Province,
South Africa. Journal of African Archaeology 11: 133–153.
Thorn, V.C. 2001. Oligocene and Early Miocene phytoliths from
CRP-2/2A and CRP-3, Victoria Land Basin, Antarctica. Terra Antarctica
Tobias, P.V. 1949. The excavation of Mwulu’s Cave, Potgietersrust
district. South African Archaeological Bulletin 4: 2–13.
Tostevin, G.B. 2013. Seeing Lithics: a Middle-Range Theory for Testing for
Cultural Transmission in the Pleistocene. Oxford: Oxbow Books.
Twiss, P.C. 1992. Predicted world distribution of C3 and C4 grass
phytoliths. In: Rapp Jr., G. & Mulholland, S.C. (eds) Phytolith System-
atics: Emerging Issues: 113–128. New York: Plenum Press.
Twiss, P.C., Suess, E. & Smith, R. 1969. Morphological classification of
grass phytoliths. Proceedings: Soil Science Society of America 33:
Van der Ryst, M.M. 1998. The Waterberg Plateau in the Norther n Prov-
ince, Republic of South Africa, in the Later Stone Age. BAR Interna-
tional Series 715: 1–158.
Van der Ryst, M.M. 2006. Seeking shelter: Later Stone Age hunters,
gatherers and fishers of Olieboomspoort in the western Waterberg,
south of the Limpopo. Unpublished PhD thesis. Johannesburg: Uni-
versity of the Witwatersrand.