Content uploaded by Sara Watson
Author content
All content in this area was uploaded by Sara Watson on Feb 21, 2020
Content may be subject to copyright.
The Doring River Archaeology Project: Approaching the Evolution of
Human Land Use Paerns in the Western Cape, South Africa
MATTHEW SHAW
Centre for Archaeological Science, School of Earth, Atmospheric and Life Sciences, University of Wollongong, Wollongong NSW 2522,
AUSTRALIA; ms152@uowmail.edu.au
CHRISTOPHER J.H. AMES
Centre for Archaeological Science, School of Earth, Atmospheric and Life Sciences, University of Wollongong, Wollongong NSW 2522,
AUSTRALIA; cames@uow.edu.au
NATASHA PHILLIPS
Centre for Archaeological Science, School of Earth, Atmospheric and Life Sciences, University of Wollongong, Wollongong NSW 2522,
AUSTRALIA; np989@uowmail.edu.au
SHERRIE CHAMBERS
Centre for Archaeological Science, School of Earth, Atmospheric and Life Sciences, University of Wollongong, Wollongong NSW 2522,
AUSTRALIA; sjrc913@uowmail.edu.au
ANTHONY DOSSETO
Centre for Archaeological Science, School of Earth, Atmospheric and Life Sciences, University of Wollongong, Wollongong NSW 2522,
AUSTRALIA; tonyd@uow.edu.au
MATTHEW DOUGLAS
College of Agricultural Science and Natural Resources, University of Nebraska-Lincoln, Lincoln, NE 68583, USA; mdouglass3@unl.edu
RON GOBLE
Luminescence Geochronology Laboratory, 1400 R St., University of Nebraska-Lincoln, Lincoln, NE 68588, USA; rgoble2@unl.edu
ZENOBIA JACOBS
Centre for Archaeological Science, School of Earth, Atmospheric and Life Sciences; and, Australian Research Council Centre of Excellence for Aus-
tralian Biodiversity and Heritage, University of Wollongong, Wollongong NSW 2522, AUSTRALIA; zenobia@uow.edu.au
BRIAN JONES
Centre for Archaeological Science, School of Earth, Atmospheric and Life Sciences, University of Wollongong, Wollongong NSW 2522,
AUSTRALIA; briangj@uow.edu.au
SAM C.-H. LIN
Centre for Archaeological Science, School of Earth, Atmospheric and Life Sciences; and, Australian Research Council Centre of Excellence for Aus-
tralian Biodiversity and Heritage, University of Wollongong, Wollongong NSW 2522, AUSTRALIA; samlin@uow.edu.au
MARIKA A. LOW
Centre for Archaeological Science, School of Earth, Atmospheric and Life Sciences, University of Wollongong, Wollongong NSW 2522,
AUSTRALIA; marika.a.low@gmail.com
JESSICA-LOUISE MCNEIL
Department of Anthropology, Harvard University, Cambridge, MA 02138, USA; jlmcneil@g.harvard.edu
SHEZANI NASOORDEEN
Centre for Archaeological Science, School of Earth, Atmospheric and Life Sciences, University of Wollongong, Wollongong NSW 2522,
AUSTRALIA; shezani@gmail.com
COREY A. O’DRISCOLL
Centre for Archaeological Science, School of Earth, Atmospheric and Life Sciences, University of Wollongong, Wollongong NSW 2522,
AUSTRALIA; corey.odriscoll@outlook.com
PaleoAnthropology 2019: 400−422. © 2019 PaleoAnthropology Society. All rights reserved. ISSN 1545-0031
doi:10.4207/PA.2019.ART138
Doring River Archaeology Project • 401
strained to between 500 ka (thousand years ago) and 250 ka
(Herries 2011; Wilkins and Chazan 2012). The Middle Stone
Age (MSA) is divided into the Early MSA (itself subdivided
into MSA1, MSA2a and MSA2b; [Volman 1981]), Still Bay,
Howiesons Poort, post-Howiesons Poort, and Late MSA.
The termination of the MSA is regionally staggered across
southern Africa (Bousman and Brink 2017; Loftus et al.
2016) ranging from 44 ka to 26 ka (Opperman 1996; Villa et
al. 2012). The Later Stone Age (LSA) is divided into Early
LSA, Robberg, Oakhurst, and Wilton units. From around 2
ka, the arrival of herders in the region signals a nal set of
technological shifts which are generally identied by the
presence of poery and the remains of domesticates which
might be classied as Neolithic (Sadr 2015).
While caves and rock shelters (hereafter rock shelters)
usually provide good preservation of the record and are
particularly useful for sequence-building, they necessar-
ily represent limited points on the landscape and likely do
not encompass the full range of prehistoric behavior. Fur-
thermore, and with a few notable exceptions (Barham 1989;
Carter 1978; Deacon 1976; Fisher et al. 2013; Hall 1990; Ma-
INTRODUCTION
The archaeological record of Africa extends across at
least 3 million years (Harmand et al. 2015; McPher-
ron et al. 2010). Within this context the southern African
sequence is particularly well resolved due to a long history
of research and the numerous rich and well-stratied sites
in the region. Though open-air sites featured heavily in
early southern African archaeology (Feilden 1884; Gooch
1882; Goodwin and van Riet Lowe 1929; Sampson 1968),
the resolution of the record which the region now enjoys is
built on the excavation of rock shelters, many of them pro-
viding sequences spanning much of the last 100,000 years
(Beaumont 1978; Carter et al. 1988; Deacon 1979; Kaplan
1990; Parkington 1980; Porraz et al. 2013; Singer 1982; Wad-
ley 1997; Wadley and Jacobs 2006; Wendt 1972).
The regional sequence as presently understood is di-
vided into three ages, each subdivided into culture-historic
units dened by distinctive aspects of lithic technology.
The Earlier Stone Age (ESA) is divided into Acheulean and
Fauresmith units and starts as early as 2 Ma (million years
ago), though the termination age is somewhat poorly con-
ROSARIA B. SAKTURA
Centre for Archaeological Science, School of Earth, Atmospheric and Life Sciences, University of Wollongong, Wollongong NSW 2522,
AUSTRALIA; rk693@uowmail.edu.au
T. ALEXANDRA SUMNER
Department of Anthropology, DePaul University, Chicago, IL 60614, USA; asumner2@depaul.edu
SARA WATSON
Department of Anthropology, University of California-Davis, Davis, CA 95616, USA; sewatson@ucdavis.edu
MANUAL WILL
Department of Early Prehistory and Quaternary Ecology, University of Tübingen, Schloss Hohentübingen, 72070 Tübingen, GERMANY;
manuel.will@uni-tuebingen.de
ALEX MACKAY
Department of Archaeology, University of Cape Town, Rondebosch 8801, SOUTH AFRICA; amackay@uow.edu.au
submied: 27 August 2019; accepted 6 October 2019
ABSTRACT
The open-air archaeology of southern Africa is extremely rich, yet has been only modestly inuential in construc-
tions of Late Pleistocene human behavior. Here we report on two seasons of work conducted as part of the Doring
River Archaeology Project, which aims to reveal paerns of human land use and technological decision-making
from the Earlier Stone Age through to the appearance of herders in southern Africa’s semi-arid interior. Across
those two seasons we have mapped and analyzed more than 20,000 cores and tools across six open-air localities,
with the small sample of available ages suggesting the accumulation of archaeologically-rich sediment bodies
along the Doring River extends back to at least 200,000 years. Our results suggest clustering of artifacts at mul-
tiple temporal and spatial scales, from individual knapping events to aggregates of hundreds of bifacial tools. All
known phases of the archaeological record appear to be represented in these assemblages, and previously docu-
mented contrasts between occupational paerns in the region’s open-air and rock shelter localities is reinforced.
These data conrm the critical importance of incorporating open-air data into depictions of the human past in
studies of the African Paleolithic.
402 • PaleoAnthropology 2019
to retain some ponded water throughout the year. Promi-
nent rivers to the west include the Groot, Driehoeks, Tra
Tra, Biedouw, and Brandewyn Rivers, with the Groot and
Biedouw in particular sustaining ows into early summer.
The Doring River itself, particularly in the middle and low-
er reaches where it cuts through Bokkeveld and Nardouw
Formation geology, is typically incised to a depth of more
than 200m below the surrounding plateaux. The resulting
steep valley through which it runs provides limited poten-
tial for signicant migration during the Quaternary.
Soils in the Doring River catchment are typically sandy,
acidic, and nutrient poor (Quick and Eckardt 2015), which
combined with low annual rainfall has restricted the inten-
sity of farming in the area. The many abandoned Europe-
an-built stone structures along the river valley aest to the
diculties of sustaining subsistence-level food production
in the area since the rst colonial loan farms were estab-
lished in the 1730s (Mitchell 2009). Nevertheless, grazing,
particularly of sheep and goats, has been a persistent fea-
ture of European land use in the Doring River valley over
the last 280 years (Neumark 1957). This is a continuation
of preceding Khoisan herding in the area (Smith and Ripp
1978), albeit that indigenous pastoralists were more mobile
which likely limited the impacts of grazing on specic loca-
tions, in contrast to the tenure-tethered grazing practices of
the later colonists.
Human use of the Doring landscape extends from the
colonial and indigenous pastoralist periods back into the
Middle Pleistocene, based on the presence of characteris-
tic ESA bifacial implements (i.e., handaxes) at the locality
Uitspankraal 1 (Bleed et al. 2017). Stone artifacts are abun-
dant on river-side sediment bodies in both upstream and
downstream locations that have been surveyed (Mackay et
al. 2014a; Smith and Ripp 1978). This abundance reects the
fact that the Doring River is not only a major regional water
source, but also an important source of raw material for
stone artifact manufacturing. The bedload of the river in-
cludes abundant cobbles of quarite and hornfels, as well
as rare cobbles of silcrete and small pebbles of chert and
quar (Low et al. 2017). Sources of hornfels are otherwise
unknown west of the Doring River, though primary horn-
fels outcrops occur along dolerite dykes on the Karoo side
of the river. The river-derived and outcrop-derived horn-
fels rocks can be dierentiated by their cortex, the former
being smooth, black, and rounded, and the laer typically
having angular planes and rugose orange-brown surfaces.
Quarite is available throughout the geology of the
broader catchment, which comprises alternating beds of
shale and sandstone/quarite in the Nardouw, Bokkeveld,
and Wieberg formations (Visser and Theron 1973). Small
quar pebbles are also commonly found eroding from the
conglomerates, particularly in the Nardouw Formation
units that prevail in the north-western (downstream) end
of the catchment. Silcrete, a ne-grained pedogenic rock,
occurs in two known primary locations in the Doring River
catchment, Swartvlei and Agtersfontein, at elevations of
465masl (metres above sea level) and 580masl respectively,
or approximately 250m and 325m, respectively, above the
zel 1989; Parkington 1988; Wadley 1984), excavation pro-
grams in southern Africa have typically focussed on single
shelters or shelter complexes, with the eect that the data
points these provide are spatially isolated from one an-
other. This approach, although providing rich, local-scale
data, makes it dicult to understand broader paerns of
land use and selement organization, and particularly how
these adaptive structures evolved during the Pleistocene
and Holocene. That objective is more fruitfully pursued
through the integration of rock shelter sequences with open
site data (Hallinan and Parkington 2017), as only by exam-
ining these archives together can we begin to understand
the breadth of land use practices in the Pleistocene record.
The long-term objective of the Doring River Archaeol-
ogy Project (DRAP) is to explore lithic technological orga-
nization as a window into the evolution of human planning
and mobility through the Late Pleistocene and Holocene.
The ability to adjust systems of movement—including the
frequency, duration, purpose, and group-composition of
moves—to changing resource congurations is a key ele-
ment of the adaptive behavior of ethnographically-docu-
mented hunter-gatherers (Binford 2001; Kelly 1995). The
distribution and form of stone artifacts across landscapes
reects decisions about stone acquisition and transport that
are expected to be sensitive to the distribution of key re-
sources mediated by systems of mobility (Andrefsky 1994;
Bamforth 1991; Nelson 1991). Thus, assemblages of stone
artifacts when studied at the landscape scale have the po-
tential to inform us of changing paerns of ancient land use
and, potentially, the evolution of this adaptive capability.
In this paper, we introduce the Doring River study
area, including the distribution of resources likely to have
inuenced decisions about subsistence movements and
technological systems, as well as the area’s known archaeo-
logical archives. We then describe the project methodology
and initial results of the rst two eld seasons in 2018 and
2019. These results conrm the abundance, antiquity, and
paerning of archaeological material in both open and rock
shelter sites in the Doring River, and highlight directions
for future work.
STUDY AREA
Situated on the eastern, rain-shadow side of the Ceder-
berg mountains, the Doring River drains approximately
28,000km2 of semi-arid shrublands at the junction of the
Fynbos and Succulent Karoo biomes (Figure 1). Rainfall in
the area is strongly seasonal; the Doring River typically be-
gins to ow following the onset of austral winter rains in
May/June, and ceases to ow by the start of the hot, dry
summer months (November/December) (Paxton 2008),
being reduced thereafter to a chain of diminishing water-
holes, some of which persist until the river begins to ow
again in the subsequent winter. Surface water is otherwise
limited in the Doring catchment. The eastern tributaries,
such as the Bos and Tankwa Rivers that drain the Tankwa
Karoo are wide, braided streams with deep, sandy channel
lls. These eastern rivers only ow after heavy rains. The
western tributaries drain the Cederberg and are more likely
Doring River Archaeology Project • 403
Figure 1. Location of the study area in southern Africa relative to modern rainfall zones. (A) Study sites and localities against regional digital elevation model (Jarvis et al.
2008). Red dots are rock shelter sites and yellow triangles are open-air localities on the Doring River discussed in this paper.
404 • PaleoAnthropology 2019
source-proximate locations such as Uitspankraal 7.
The Early LSA assemblage from Uitspankraal 7 was re-
covered from a surface cluster of technologically coherent
archaeological material in one small area of the 40,000m2
sediment stack. This was one of several clusters identied
at the locality during work in 2014 (Low et al. 2017; Will et
al. 2015). Material apparently associated with the Late Ho-
locene with Neolithic, and mid Holocene Wilton, Oakhurst,
Early LSA, post-Howiesons Poort, and Still Bay all showed
distinct paerns of distribution and clustering across the
surface. The paerning of this surface material prompts
several questions: 1) Was the material exposed on the sur-
face since the time of original deposition, or was it buried
after discard and subsequently exposed by erosional pro-
cesses? 2) If buried and then exposed, how long would
surface-exposed material take to disaggregate—that is, to
lose its paern of clustering—under modern erosional con-
ditions? 3) Furthermore, what distortions in assemblage
composition would result from this process? That is, were
certain elements more likely to be dispersed than others,
and how might this aect interpretations of the archaeo-
logical data?
To investigate these questions, we conducted an ac-
tualistic experiment with a replicated assemblage that
mimicked a typical LSA microlithic assemblage, compris-
ing freehand and bipolar cores, as well as small akes and
blades (Phillips et al. 2018). We placed and mapped the as-
semblage in 2014, and regularly documented spatial disag-
gregation across 22 months. Many of the artifacts moved
considerable distances within the observation period, to the
extent that disaggregation of the assemblage would have
occurred within a few centuries, and that within a millen-
nium most of the assemblage would have been incorpo-
rated into the bedload of the Doring River. Given this, per-
sistent or sustained subaerial exposure of the archaeology
of the Doring River sediment stacks throughout the past is
expected to have erased the kinds of the spatial paerning
we had previously documented, and which we document
further below. Alternatively, brief or intermient periods
of exposure and/or net sediment loss in the past may have
redistributed artifacts from certain periods while preserv-
ing paerning in others. While these and other formation-
al/erosional models remain to be tested, it is clear that the
sediment stacks are heavily erosional under present condi-
tions, a point reinforced by under-cuing of historical Eu-
ropean structures on the sediment bodies (see below).
The most likely cause of recent erosion is grazing of
stock under European land tenure (Smith and Ripp 1978).
Due to the limited water and poor feed in the Doring catch-
ment, stock would inevitably have been concentrated in and
around the river for long periods, accelerating already high
natural rates of erosion (Phillips in preparation). The eect
of grazing on these sediment stacks is particularly evident
at the locality of Klein Hoek 1 where a fence line provides
a comparison of a pre and post grazing area (mentioned
in the results sections) These sets of observations provided
impetus for the Doring River Archaeology Project.
river. Swartvlei is characterized as a weathered rock man-
tled surface, consisting of ne to medium-grained textured
silcrete with very well- to moderately-well sorted clasts.
Agtersfontein is characterized by three outcrops and sur-
rounding rock mantled surface. Material ranges in texture
from ne-grained, well sorted clasts to very poorly sorted
with pebble-sized quar inclusions. Our surveys for sil-
crete in the catchment are currently incomplete, however,
and more sources may be discovered. Based on visual
classication, material from both known sources has been
identied in the assemblages discussed below.
PREVIOUS ARCHAEOLOGICAL WORK
Six rock shelters have been excavated in the catchment
of the Doring River, though we concentrate here on the
ve located downstream of the Groot/Doring conuence:
Hollow Rock Shelter, Klipfonteinrand 1, Klipfonteinrand
2, Mertenhof, and Putslaagte 8 (see Figure 1). These sites
are located at 2km (Putslaagte 8), 13km (Klipfonteinrand
1 and 2), 17km (Hollow Rock Shelter), and 19km (Merten-
hof) from the Doring River, allowing us to explore paerns
of material transport away from the major regional water
source. The sequences from these sites are reasonably co-
herent (Low and Mackay 2016; Mackay et al. 2015) and
consistent with generally recognized paerns of regional
technological change (Table 1). The only notable, recurrent
weakness in these sequences across the last ~80 kyr is the
limited amount of material found from the early to mid-
Holocene, though artifact densities are typically also very
low throughout the Late MSA (~50–25 ka).
In addition to these rock shelters, a single open site has
been excavated on the Doring River at Putslaagte 1 (Mack-
ay et al. 2014b). The locality is one of several discrete sedi-
ment stacks which occur along the Doring River corridor,
and which invariably preserve stone artifact assemblages
from the Earlier, Middle, and Later Stone Ages. The large
assemblage recovered from Putslaagte 1 dates to <58 ka and
has been assigned to the Late MSA. The technological sys-
tem found at the site is distinctive, and features reduction
of at hornfels cobbles using simple prepared platforms to
produce large and usually cortical akes. Platforms on the
Putslaagte 1 cores were rarely re-prepared and cortex ratio
analysis suggests that the cortical akes were transported
away from site, presumably for use elsewhere in the catch-
ment (Lin et al. 2016; see also Holdaway and Douglass 2015
for discussion of approach).
Similar transport paerns have been inferred based
on comparison of the Early Later Stone Age sample from
Putslaagte 8 with that from another sediment stack local-
ity, Uitspankraal 7 (Low et al. 2017). Here again, domi-
nant aking systems make use of hornfels cobbles, in this
case to produce blades exploiting natural ridges along the
cobble edge. Both blades and blade cores are abundant at
Uitspankraal 7, but only blades appear in signicant num-
bers in the Putslaagte 8 Early LSA assemblage, which is
located 2km from the Doring River and thus the source of
hornfels. It is inferred that the hornfels blades were prefer-
entially transported, whereas the cores were discarded at
Doring River Archaeology Project • 405
on the inner portion of major river bends. Due to its deeply
incised prole, accommodation space for long term accu-
mulation of sediment is generally limited along the Doring
River, such that there are no terrace successions. The sedi-
ment stacks are thus relatively easily dierentiated from
modern overbank deposits by their elevation (3–30m above
the river), distance from the nearest active river channel
(usually >100m), distinctive reddish orange color (the mod-
ern river sands are white), and indurated texture. Due to
the low frequency of suitable accommodation space, and
the intensely erosional nature of the region, the stacks tend
to have well-dened boundaries at which the underlying
colluvium is exposed.
Nominally, identication of sediment stacks localities
would occur as part of Phase I, and specically during re-
connaissance surveys that take in the length of the Doring
River from its conuence with the Bos River to the point
where it merges with the Olifants River. That reconnais-
sance would involve both mapping and analysis of rele-
vant artifacts (see below) on sediment stacks, and also any
material encountered during walking between sediment
stacks—thus characterizing the background scaer of ma-
terial along the river. In practice, across the rst two sea-
sons we only focussed on stacks that had been identied
during prior work in the area (Mackay et al. 2014b), with no
‘o-stack’ survey or analysis undertaken as yet. Such work
will become more common as we move into the less well-
surveyed upper and lower reaches of the river.
PHASE I
Once a sediment stack has been identied, its boundaries
and major sediment units are mapped. Thereafter, the goal
of Phase I analysis is rapid appraisal of the artifact distribu-
tions with sucient information to make preliminary inter-
pretations of any clustering that might inform Phase II and
III research priorities (see below). As such, Phase I has two
components. The rst component involves the recording
and analysis of all cores, implements (i.e., retouched stone
tools), anthropogenically modied non-aked stone imple-
ments, ochre, poery, and historic metal and glass across
all identied sedimentary stacks with no size cut-o (SOM
Table 1). All anthropogenically modied ochre was record-
AIMS AND METHODS
The aims of this project are inuenced by the sensitivity
and fragility of the sediment stacks as well as the overall
scale of the Doring River catchment. A method that allows
for the expeditious documentation of high-resolution data
was implemented as a Phase I approach, focusing on the
recording of specic artifact types (implements, cores, and
some non-aked objects) that would serve as the founda-
tion for further decisions on future analysis (Phases II and
III). The aims of the Doring River Archaeology Project are
thus as follows:
1. to record the distribution of archaeological material
across sediment stacks along the Doring River;
2. to assign temporal ranges where possible to this ma-
terial both by relating technological characteristics to
those from the known regional sequence, and by ra-
diometric dating where possible;
3. to examine similarities and dierences between the
composition and technological characteristics of as-
semblages at localities along the river with those re-
covered from rock shelter sites 2km, 13km, 15km and
19km away; and,
4. to use the information from 1, 2, and 3 above to un-
derstand past paerns of land use and technological
planning in the catchment.
The project has three methodological phases that can
run separately or concurrently. While we will outline the
methods for all phases, here we concentrate on Phase I as
this accounts for the bulk of work completed so far, and for
the results that we will later present. Prior to describing the
phases of data collection, however, we outline the process-
es by which sediment stack localities1 were identied and
dened, and protocols for mapping ‘o-stack’ material.
IDENTIFICATION OF ARCHAEOLOGICAL
LOCALITIES
Our analyses so far have concentrated on the distribution
of archaeological material as it occurs on sediment stacks
along the Doring River. As noted earlier, these sediment
stacks are large, discrete accumulations of sandy sediment
that occur in low energy contexts along the river—typically
at the conuence of the Doring and its minor tributaries, or
TABLE 1. LOCAL ROCK SHELTER SEQUENCE SUMMARY BASED ON SITES SHOWN IN FIGURE 1.
Locality
ESA
Early MSA
Still Bay
HP
Post-HP
Late MSA
Early LSA
Robberg
Oakhurst
Early/mid
Holocene
Wilton
Pottery
PL8
>76
?
71-58
71-58
33-45
25-21
21-18
<18
<1
KFR1
pres.
pres.
pres.
22-16
16-13
10.1-6.3
6.3-3.8
KFR2
4.7-3.6
1.7-1.3
HRS
?
87-72
MRS
pres.
pres.
pres.
pres.
pres.
pres.
?
pres.
406 • PaleoAnthropology 2019
corded could be assigned to culture historic units. This is
partly because many of the artifacts were in the early stages
of reduction—unsurprising given that the Doring River is a
source of both hornfels and quarite for artifact manufac-
ture—and partly due to issues of identiability discussed
at greater length below. Essentially, we have to accept a
high rate of Type II (false negative) errors in our culture
historic assignment during Phase I analysis, due to the high
proportion of lithic artifacts that transcend named stone
tool technocomplexes (Shea 2014). Type I errors (false posi-
tive) are likely to be less common due to limited overlap in
‘diagnostic’ artifact types but will have occurred, especially
because analysts were encouraged to make assignments to
culture historic units where probable with reference to the
excavated collections; some of these aributions will inevi-
tably have been erroneous. At least one principal analyst
with extensive experience in the local sequence was pres-
ent throughout the surveys to help limit such false assign-
ments.
In addition to in-eld analysis, artifacts assigned to cul-
ture historic units were 3D scanned using a Rexscan DS2
Silver structured light scanner. Distinctive implement and
core types were also scanned at the discretion of the ana-
lyst. During 2018, artifacts were scanned in-eld, the scan-
ner being powered by a dual-baery system installed in
the project vehicle. Due to high temperatures at the start of
Season 2, scanning was conducted at the eld house. Arti-
facts to be scanned had their locations recorded with a met-
al tag aached to a nail, were scanned the next day at the
eld house, and then returned to their point of origin. The
same protocol was used for detailed artifact photography
of selected artifacts. The scans are intended to serve several
purposes. The rst is for visual communication of key arti-
fact types. The second is to enable comparison of artifacts
between open-air localities, and between open-air localities
and excavated collections. The third is as an archive—as
noted above, the open-air localities of the Doring River ap-
pear to be eroding quickly under modern conditions, and
many of the artifacts that are currently on the surface will
conceivably be lost within a generation. After two seasons,
1521 artifacts have been scanned as part of Phase I.
The artifact aributes collected as part of individual
point data allow large clusters of similar artifacts to be
identied, whether or not these can be assigned to culture
historic units. However, small clusters can be dicult to
identify, and many clusters are poorly represented by cores
and implements, occurring instead as concentrations of
akes. In order to control for this, the second component of
Phase I involved mapping clusters of artifacts as polygons.
Diverse criteria were allowed when deciding whether to
map a cluster with a polygon. Polygons were made wher-
ever a spatially-coherent assemblage of technologically
similar artifacts was observed. They were also made where
clusters of artifacts of a similar, distinctive material were
observed. Finally, they were used to map scaers of ret
sets involving three or more artifacts. While no dedicated
reing work was undertaken, opportunistic rets were re-
corded when observed.
ed, as well as unmodied fragments greater than 30mm in
maximum dimension; no other artifact classes were size-
limited for analysis. The decision to focus on cores and im-
plements for Phase I reects their potential to act as time-
sensitive markers relative to the known regional excavated
sequence. The decision is also pragmatic—the sheer quan-
tity of unretouched akes and other fragments, estimated
to be in the hundreds of thousands per locality—would
make analysis of even a limited proportion of them hugely
time-consuming, limiting our potential for coverage of the
archaeology of the catchment. Thus, though akes can, of
course, be equally if not more informative about time-spe-
cic technological behavior, their analysis is restricted to
Phase II.
No artifacts were collected during Phase I—all were
analyzed in-eld and replaced at their point of origin. The
spatial location of each artifact was recorded using a hand-
held mobile GIS system. In Season 1 (2018), artifacts were
recorded on Trimble Juno 3B units using ArcPad 10.2 and
the device’s internal receiver, with an accuracy of 3–5m.
Analysis in 2018 was conducted concurrently by two lithic
analysts. In 2019, artifacts were recorded using the ESRI
ArcGIS Collector mobile application on Apple iPad Mini
4’s linked wirelessly by Bluetooth to a Bad Elf Surveyor Pro
GNSS receiver with an accuracy between 1–3m (see Ames
et al. submied for methodological details). The shift to
this system allowed us to incorporate Bluetooth-enabled
Sylvac Cal-evo digital calipers, providing the capability for
rapid and low-error capture of a limited set of metric data.
The internal cameras on these devices also allowed pho-
tographs to be taken of most artifacts (minimal cores and
unworked ochre were typically not photographed). Full,
non-overlapping coverage of each locality was achieved
by dividing each locality into parallel ~2m-wide transects
using black nylon string. The improved recording system
(see Ames et al. submied) made it possible to have more
concurrent lithic analysts (faculty, postdoctoral fellows,
and post-graduate students), who were shadowed by ar-
chaeological student volunteers from universities in South
Africa and Australia.
A limited set of traits was recorded for all artifacts,
summarized in SOM Table 1. Where possible, artifacts
were assigned to culture historic units as established in the
prevailing regional framework augmented with data from
excavated sites in the catchment (for aributes and SOM
Table 2 for time-sensitive artifacts). The functional assump-
tion here is that certain kinds of artifacts are more common
during certain time periods than others, and can thus act
as pseudo-markers for those periods (though see below).
To be clear, this approach is simply a preliminary means
for estimating the age of archaeological material in what
are surface scaers which we intend to test and ideally
replace once we have excavated samples and radiometric
ages. To facilitate consistency between researchers, a refer-
ence collection of artifacts from excavations at Putslaagte 1,
Putslaagte 8, Klipfonteinrand, and Mertenhof was put on
display in the eld house.
Inevitably, only a proportion (7.2%) of the artifacts re-
Doring River Archaeology Project • 407
and open sites has been noted towards the west coast of
South Africa through the mid to Late Holocene (Jerardino
and Yates 1996), a period that is also poorly represented in
rock shelters in the Doring River catchment. In spite of this,
the regional framework is almost exclusively constructed
using rock shelter data.
Second, it is not reasonable to assume that technologi-
cal behaviors at dierent points on the landscape will al-
ways resemble each other, even if they occur in the same
period. While the culture historic framework provides an
averaged depiction of technology in a spatial block at a
period of time, decisions about what kinds of artifacts to
transport, in addition to potential functional dierences
between sites, may result in partitioning of a technologi-
cal system at dierent locations (Barton and Riel-Salvatore
2014). Previous work in the Doring River catchment sug-
gests the operation of such factors, such that, for example,
Late MSA and the Early LSA assemblages identied in one
part of the system are constituted very dierently from the
expression of the same system elsewhere in the catchment.
Third, even allowing for clustering of temporally-co-
herent artifacts, exposed open-air site assemblages are al-
ways prone to conation of occupation from multiple pe-
riods—the palimpsest eect (Bailey 2007)—and these can
be dicult if not impossible to disentangle. Thus, we can-
not expect the signal in our Phase II analyses to be entirely
‘clean’ to the extent that that term is meaningful.
Recognizing these limitations, the project incorporates
excavation and radiometric dating of sediment units ad-
jacent to targeted clusters, including both those for which
we presume already know the age, and those for which
we have no analogues in the regional framework. Phase III
work will commence once Phase I has been completed and
we have a more comprehensive understanding of the avail-
able resources across the studied open-air localities. Exca-
vation will ultimately facilitate our project aim of under-
standing technological behavior across space and time in
the Doring River catchment, with culture history providing
a functional, if limited, interim framework.
RESULTS
At the completion of two seasons of work, mapping of arti-
fact distributions across six sediment stacks has been com-
pleted. The total analyzed sample so far is 24,221 artifacts,
comprising 17,646 cores, 3657 retouched akes, and 2918
other pieces (including poery, ochre, anvils, grindstones
and hammerstones). Here we focus our aention on ve
stacks that illustrate the range of depositional contexts,
ages, and clustering paerns that occur. Artifact summaries
for these localities are given in Table 2 (raw material), Ta-
ble 3 (implement typologies), and Table 4 (industries). We
describe the completed stacks in a downstream sequence
starting at the Biedouw/Doring conuence.
UITSPANKRAAL 9 (UPK9)
Situated at the conuence of the Doring and Biedouw Riv-
ers, Uitspankraal 9 (pronounced ate-spun-krahl) (UPK9) is a
low vegetated rise at the distal end of a colluvial slope, cov-
All sediment stacks analyzed during Phase I were pho-
tographed using a DJI Mavic Pro drone (1/2.3” 12 Mega-
pixel censor), which was own over the locality in a grid
paern at 40m altitude. The resulting images were stitched
in AgiSoft Photoscan and used to generate high-resolution
orthomosaics, and further processed to create vegetation-
free digital surface models of each stack. These surface
models can in turn be used to map ow paths and estimate
erosional sensitivities (Ames et al. submied; Howland et
al. 2018). It is important to recognize that there are com-
plex issues with creating UAV derived surface models that
require explicit treatment before they can be used in a sci-
entically valid framework; an issue we consider in more
detail elsewhere (see Ames et al. submied). The boundar-
ies of each sediment stack, as well as the major sedimentary
units within them were mapped using the same mobile GIS
platform used for recording artifact polygons.
PHASE II
Phase II involves in-eld analysis of all artifacts (including
unretouched akes and fragments) over 20mm, in targeted
clusters identied during Phase I, and their mapping in situ
using a Trimble R7/R8 Real Time Kinematic (RTK) GNSS
system. Base station coordinates are post-processed using
the Canadian Geodetic Survey of Natural Resources online
precise point positioning service, which were then used to
recompute the RTK rover datasets in Trimble Geomatics
Oce Version 1.63. Whereas the objective for Phase I is to
identify sediment bodies and characterize the broader spa-
tial and technological paerns on these surfaces, the objec-
tive of Phase II is to provide the ne-grained technological
data necessary to understand lithic transport and reduction
through the catchment—by making comparisons both be-
tween Doring River assemblages and between the river as-
semblages and those from the excavated rock shelters.
Clusters for Phase II are thus selected based on the
information they are likely to oer concerning particu-
lar technological systems or temporal intervals relative to
previously analyzed samples. For example, the Phase II
analysis of a Robberg cluster at the locality Uitspankraal
9 undertaken by Sara Watson in 2018 was intended to test
propositions by Low and Mackay (2018) regarding the
characteristics of Robberg technology near sources of raw
material. So far, Phase II analysis has been completed on
4583 artifacts from two clusters, though the results are not
presented here.
PHASE III
As suggested in the description of the Phase I methodol-
ogy, there are problems with aempting to relate sets of
artifacts from open site contexts to those from the local and
regional culture-historic framework. First, the framework
is incomplete. A strong example is the regional paucity
of archaeological evidence from the Late MSA (Mitchell
2008). The assemblage from Putslaagte 1 suggests that
this paern may reect limited use of rock shelters in this
period rather than limited presence in the region. Similar
paerns of alternation between occupation of rock shelters
408 • PaleoAnthropology 2019
TABLE 2. LITHIC RAW MATERIALS, PROPORTION BY LOCALITY FOR MAJOR LITHOLOGIES ONLY.
Locality
Quartzite
Hornfels
Quartz
Silcrete
Chert
Sandstone
UPK9
39.5%
28.4%
13.0%
5.1%
12.2%
1.7%
UPK7
39.4%
38.2%
8.9%
6.4%
4.6%
2.5%
UPK1
51.4%
38.5%
4.3%
2.5%
2.1%
1.2%
KH1
36.4%
53.7%
2.5%
2.8%
3.0%
1.5%
DB8
17.8%
67.3%
5.5%
4.3%
4.7%
0.2%
PL8
27.5%
64.9%
0.8%
3.9%
1.2%
1.7%
TABLE 3. COUNTS FOR MAJOR IMPLEMENT CLASSES BY LOCALITY.
Locality
Backed
Denticulate
Grindstone
Hammerstone
Naturally
Backed Knives
Point-bifacial
Point-unifacial
Scaled piece
Scraper-other*
UPK9
24
13
51
259
102
1
2
121
823
UPK7
12
34
36
57
13
12
17
109
174
UPK1
3
1
5
6
0
8
3
6
40
KH1
3
12
3
30
3
133
12
10
82
DB8
8
4
9
20
6
2
11
18
61
PL1
0
0
1
3
0
0
0
8
2
*The class ‘scraper-other’ includes a wide range of scraper forms such as lateral scrapers, end
scrapers, and continuous or thumbnail scrapers.
TABLE 4. COUNTS FOR ARTIFACTS ASSIGNED TO INDUSTRIES BY LOCALITY.*
Locality
ESA
Early MSA
Still Bay
HP
Post-HP
Late MSA
Early LSA
Robberg
Oakhurst
Early/mid Holocene
Wilton
Pottery
UPK9
1
?
1
5
2
12
1
267
238
64
83
509
UPK7
0
?
19
11
51
95
55
36
59
?
49
178
UPK1
29
?
9
5
2
28
1
0
0
?
4
105
KH1
3
?
183
10
32
230
10
28
0
?
10
0
DB8
0
?
6
6
10
28
1
20
6
?
9
2
PL1
0
?
0
0
2
81
4
0
0
?
0
0
Total
33
?
218
37
99
474
72
351
303
?
155
794
*Question marks used for the Early MSA and Early to Mid-Holocene reflect the lack of clear identifying
characteristics for these periods. Red-bold is used to denote localities at which a given industry occurs in a
cluster with apparently intact flaking debris, i.e., not only the diagnostic cores and implements. Questions
marks and light grey shading indicate industries for which we currently lack appropriate diagnostic markers.
Doring River Archaeology Project • 409
the locality are required to test this proposition.
The eastern part of UPK9 features more than 200 arti-
facts assigned to the Robberg (22–16 ka), which are com-
paratively rare in the western area. Robberg artifacts oc-
cur in three clusters—two dominated by silcrete and one
by chert—which fringe the margins of the intact sediments
but are largely absent from the erosional features on top
of them. The general distribution of Robberg artifacts sug-
gests that they either underlie the sediment stack in this
area or occur close to the contact between the calcrete and
the lowermost sands. A Phase II analysis has been conduct-
ed on one of the silcrete-rich Robberg clusters in this area
with a publication in preparation.
Oakhurst artifacts are prolic in the lag at the north-
ern edge of this unit. The cluster here includes 80 naturally
backed knives (see Figure 2) and 45 large scaled pieces.
Both of these artifact types are made on hornfels, however,
the laer are exclusively made on hornfels derived from
the river cobbles (based on the cobble morphology of the
cortex). The naturally backed knives, on the other hand, are
made exclusively on hornfels derived from primary sourc-
es—contact metamorphism of clay-rick sediment adjacent
to intrusions, such as dolerite dykes in the interior Karoo—
identiable by its characteristically rough orange exterior
and near 90° joints. A small number of Wilton artifacts oc-
cur in pockets of exposure towards the top of the stack. Pot-
tery is distributed in clusters across the entire area, as are
historical artifacts including a saddle badge with an 1851
date, suggesting occupation through to the near-present.
Not all of the clusters that we identied can easily be
assigned to culture-historic units. Quar is generally un-
common in the assemblages of the Doring River (Low et
al. 2017; Mackay et al. in press), however, the UPK9 as-
semblage reveals a striking concentration of bipolar cores
made on this material towards the centre of the locality (see
Figure 2). The area is adjacent to, but not overlapping, a
cluster of Robberg-associated artifacts, and lacks other time
markers that we can identify. Poery is common around
the concentration but no more so than elsewhere in the
eastern stack. Quar bipolar cores are the dominant signal
in the Late Holocene assemblages at Klipfonteinrand 2 and
Putslaagte 8, however, poery also occurs in those layers
(Nackerdien 1989; Plaske 2012); currently this is the most
plausible assignment for these artifacts, but further work is
required to clarify this suggestion. Assemblages from ear-
lier parts of the Holocene in this region typically show a
predominance of scrapers (Thackeray 1977), which are not
in evidence in this cluster.
Chert is similarly uncommon in regional assemblages,
but again displays strong clustering at UPK9. As noted, one
of these clusters is associated with the Robberg at the north-
ern edge of the locality. More striking is the concentration
of chert scrapers (n=64) at the south-western edge of the
major Oakhurst cluster (see Figure 2A). These artifacts are
materially and physically consistent, with heavy stepped
retouch on the lateral margins and a straight to rounded
distal edge. The most comparable artifacts we know from
the broader region are those referred to as ‘Woodlot’ scrap-
ering 27,013m2. The sediment stack here is generally <1.5m
high and sits at an elevation of 15–30m above the Doring
River. At the west end, this sediment stack has several ero-
sional embayments on both the north and south sides, the
laer being the most signicant. To the northeast, the sedi-
ments have been largely washed away, leaving a lag de-
posit of artifacts intermixed into the colluvium. The intact
eastern part of the stack is cut at the south edge by an exca-
vated feature probably constructed to constrain water run-
o along the adjacent road. This cuing reveals a ve-part
sedimentary sequence comprising basal colluvium overlain
by a well-developed nodular calcrete, another colluvium,
indurated sands, and nally loose vegetated sands that are
likely the result of recent aeolian processes. Immediately to
the south of the cuing, sediment has been eroded down to
the upper colluvium, and in the process has under-cut the
foundations of an historical structure by 400mm (Figure
2B). Two samples of nodular calcrete were excavated from
the immediate subsurface in this cuing and submied to
the Wollongong Isotope Geochronology Lab; they provid-
ed U/Th isochron ages of 226±25 ka (S91090) and 202±48
ka (S91091) (SOM Table 3). Excess detrital thorium in both
samples limits more rened ages.
The survey at UPK9 concentrated on the areas of extant
sandy sediment, with a ve meter buer into the surround-
ing colluvium. A total of 9486 artifacts was recorded in
surveys at this locality. ESA and MSA artifacts are largely
restricted to the exposed colluvium (see Figure 2), though
there is lile evidence for a lag of LSA artifacts on that unit,
suggesting both that the current extent of the sand units is
close to their original extent, and that the sand units were
the focus of occupation. The exception is the aforemen-
tioned lag at the north-eastern edge of the locality. Other-
wise, LSA and Neolithic (poery) artifacts are restricted to
exposures on the sediment stack.
The distribution of identiably time-specic artifacts
displays comparable signals of clustering to those noted
during previous work at Uitspankraal 7 (UPK7). In the
exposures on the western part of the stack, we observe a
three-part horizontal ‘sequence’ running from west to
east. At the western edge of the southern exposure, scaled
pieces, large ‘D-shaped’ scrapers known locally as natu-
rally backed knives (see further discussion below), and
core-scraper-anvils typical of the local Oakhurst (14–8 ka)
are common (see Figure 2). Moving east, we nd 17 small
‘thumbnail’ scrapers made on silcrete and chert, and sev-
eral backed artifacts, all typical of the mid-Holocene Wilton
(8–2 ka) (Thackeray 1977). Towards the end of the expo-
sure, poery characteristic of the last 2000 years appears,
and extends further into the upper exposures on the north
side of the stack. The ‘sequence’ here is interesting insofar
as the dierent components are not directly overprinted on
one another. If the artifacts were eroding from a vertically
stacked sediment sequence, then we would expect a more
typical palimpsest eect. We hypothesize that the results
instead reect a shifting focus of occupation that tracks the
eastward migration of the dune crest up the drainage de-
pression through the Holocene. Excavation and dating at
410 • PaleoAnthropology 2019
Figure 2. Distribution of artifacts and sediment bodies across Uitspankraal 9 (UPK9). (A) A sample of Woodlot scrapers from the eastern cluster. (B) Erosion beneath an
historic structure.
Doring River Archaeology Project • 411
diately to the south. On the younger sediment units in the
southern and western part of the locality, artifact density is
persistently low.
Leaving aside the Early MSA, UPK7 retains a signal
from all identiable culture-historic units. Still Bay arti-
facts are clustered on indurated red sediments in the south
eastern erosional embayment, with a few more towards the
center of the locality. A few Howiesons Poort artifacts also
occur in this central area. Post-Howiesons Poort artifacts
display strong clustering in the central area, with Late MSA
artifacts well represented across the same sediment unit,
extending to the west. The Early LSA cluster analyzed by
Low in 2014–2015 is clearly identiable in our Phase I work,
occurring as a limited blow-out in the modern dune sands
and resting on the partially consolidated yellow sandy unit
dated elsewhere to around 30 ka. These Early LSA artifacts
take up the entirety of the blowout and likely extend across
a larger area below the modern sands. Robberg artifacts are
much less common at UPK7 (n=36) than at UPK9 (n=267),
but cluster strongly at the top of the stack. Oakhurst arti-
facts are well-represented in four dierent areas, but most
notably at the top of the stack. Wilton artifacts are also con-
centrated in that area, as are potsherds, though the laer
also occur in a series of clusters in loose sands around the
fringe of the partly consolidated sands.
As with UPK9, there are clusters of material at UPK7
that we cannot easily reconcile with the known regional
sequence. While at UPK9 these typically reected behav-
ioral aggregates, some of the occurrences at UPK7 more
likely represent brief events. One of the more interesting
is a splay of quar artifacts on the southern slope, adjacent
to the OSL sample UNL-3809, thus siing above a sedi-
ment unit dating 30.3±1.3 ka (see Figure 3A). The cluster
here comprises 29 cores—of which 23 are bipolar—an an-
vil and a hammerstone are within 46m2. Twenty-two of the
cores are made from quar (all bipolar), ve from horn-
fels and one each from chert and quarite. While no re-
ts were aempted on the bipolar quar cores, one of the
hornfels cores on the southern side of the splay rets an
adjacent ake and a cortical hornfels blade located on the
other side of the cluster about 6m away. While this clus-
ter is likely LSA, and potentially quite recent, its position
on the partly consolidated sands raises the possibility that
it is coeval with the Early LSA cluster reported by Low et
al. (2017). One of the observations made in that paper was
that in rock shelter samples, hornfels blade production and
small quar bipolar aking were intermixed, but in the
open-air example, extensive hornfels blade production oc-
curred without any signicant bipolar aking—quar or
otherwise—with the implication that those components of
the technological system may have been undertaken sepa-
rately. It is thus possible, though by no means certain, that
the quar bipolar splay on the lower south-west slope at
UPK7 represents the other component of that system op-
erationalized separately.
Another ‘event’ scale cluster occurs on the northern
side of the locality, eroding out from the partly consoli-
dated sands near OSL sample UNL-3810, dating 30.5±1.4
ers dating 9–7 ka from sites in the southern Cape and Le-
sotho (Deacon 1984; Mitchell 2000); such artifacts are either
not present in the mid-Holocene samples from Klipfontein-
rand 1, or not dierentiated in Thackeray’s (1977) analysis
of it.
UITSPANKRAAL 7 (UPK7)
UPK7 is an extensive sediment stack located on an ancient
cobble terrace of the Doring, 12–26m above the current riv-
er valley and extending over the adjacent slope. The stack
itself covers 42,326m2. At the eastern edge of the stack is the
modern dune crest, with deep erosional rills incising the
sediment on the western and southern sides and exposing
the indurated lower red sediments. Like UPK9, UPK7 has a
basal colluvium but lacks an overlying calcrete. Indurated
red sediments with nodular calcrete form the oldest iden-
tiable unit, and this is overlain by a partly consolidated
yellow sand unit, and capped by unconsolidated dune
sands (Figure 3). We have so far obtained indicative mul-
tiple-aliquot optically stimulated luminescence (OSL) ages
of quar for the two upper sediment units only; results for
the lower units remain in preparation.
Samples UNL-3808, UNL-3809, and UNL-3810 were
collected from shovel-cut sections at depths of between 0.3–
0.7m below present surface (see SOM Table 3). The samples
traverse the site from east to west, though are, as noted,
restricted to the less consolidated units. Samples were ana-
lyzed at the Luminescence Geochronology Lab at the Uni-
versity of Nebraska, Lincoln, using methods described in
the SOM. Samples UNL-3809 and UNL-3810 were collected
from the partly consolidated yellow sand unit and returned
near identical ages of 30.3±1.3 ka and 30.5±1.4 ka respective-
ly, reecting accumulation of much of the sediment across
this site around the MIS 3/2 boundary. Sample UNL-3808
was taken from the overlying unconsolidated dune sands
and is dated to 0.069±0.005 ka. This suggests that the upper
dune sand stabilized in the last century, which may reect
the elevated rates of recent erosion discussed previously.
We have previously published the results of detailed
analysis of post-Howiesons Poort and Early LSA clusters
at UPK7 conducted in 2014 and 2015 (Low et al. 2017; Will
et al. 2015). Here we provide the broader context for those
clusters from our recent Phase I survey. Phase I survey re-
corded 4285 artifacts at UPK7. Unlike UPK9, MSA artifacts
are more common than those from the LSA at this local-
ity. LSA artifacts occur principally at the top of the sedi-
ment stack and on the younger sedimentary units on the
lower southern drape—those dating to around 30 ka. In the
center of the locality, where younger sediments have been
lost and where erosion is exposing and actively destroying
the indurated red sediments, MSA artifacts are dominant
and LSA artifacts essentially absent. Artifact density varies
across the locality in ways not entirely controlled by sur-
face exposure; artifacts are more commonly recorded on
the exposed indurated surfaces of these sediment bodies,
while large areas of loose sediment (younger sands) are de-
void of artifacts. Density is highest in the exposure at the
top of the stack, and on the indurated red sediments imme-
412 • PaleoAnthropology 2019
Figure 3. Distribution of artifacts and sediment bodies across Uitspankraal 7 (UPK7). (A) Cluster of quar bipolar cores beneath an anvil. (B) Silcrete ret set.
Doring River Archaeology Project • 413
present, though there is a small cluster in the upper loose
sands towards the center of UPK1.
The handaxes in the ESA samples are extremely di-
verse in size and shape, as we have noted previously, and
the production systems relatively simple (Bleed et al. 2017;
Magnani et al. 2016). The smallest complete example we
recorded in Phase I had a maximum dimension of only
80.4mm; the largest measured 217.0mm. Artifacts from the
Still Bay, Howiesons Poort, and post-Howiesons Poort—
all indicative of MIS 4 and early MIS 3—are present in
small numbers in the major erosional embayments associ-
ated with the MSA. Late MSA-assigned artifacts occur here
too but these also occur in small numbers in blowouts on
the indurated yellow sands to the west. In one blowout, a
prepared hornfels core with a single platform was ret to
two late ake removals. Given that the artifacts are almost
certainly MSA, it suggests that the lightly indurated yel-
low sands began to accumulate around or before 30 ka,
and may be consistent in age with the similar sedimentary
unit at UPK7. Across the rest of that area, the only culture-
historic unit represented is the Wilton, indicated by three
thumbnail scrapers.
Not assigned to any culture-historic unit is a distinct
concentration of akes and cores in the southern part of the
locality (see Figure 4A). All of the artifacts in this cluster
are made from ne and homogeneous grey and blue-grey
quarite available in blocks on the adjacent scree slope.
The cores are predominantly recurrent Levallois, some
with large blade removals. Large blades are also reason-
ably common elsewhere in this erosional embayment. As
with the UPK7 examples, these artifacts are inferred to be
either Late or Early MSA.
KLEIN HOEK 1 (KH1)
The locality Klein Hoek 1 (KH1) occurs at the western
(downstream) end of an extensive point bar. A fence line
denes the eastern boundary of the locality. While artifacts
are abundant in erosional features west of the fence line,
less intensive grazing to the east means that there has been
lile erosion and thus limited surface exposure of artifacts.
Artifacts no doubt exist in subsurface contexts in that area,
but for the present we dene the limits of the locality based
on the visible extent of the archaeology, an area of 19,432
m2 (Figure 5) that is 9–17m above the river.
Unlike UPK7 and UPK9, KH1 has a colluvial drape
over most of its surface. While now separated from the ad-
jacent scree slope by a minor drainage channel, we infer
that this slope was connected to the scree in the past pro-
viding a continuous colluvial surface. Artifacts throughout
the western part of the locality occur as a lag within this
colluvium. Farther to the east, sandy sediments are pre-
served and a three-part sequence can be identied with the
underlying colluvium exposed in patches throughout. The
oldest unit, a compact brownish-red sandy deposit with a
crumbly appearance and slightly friable consistence, sits
immediately on top of the colluvium and is visible in only a
few isolated areas. Covering this unit are indurated yellow-
brown sands. Areas where this unit is intact have created
ka. Here, in an area with a very low density of nds, we
identied a cluster of silcrete akes that is largely invisible
with our Phase I point data. Three of the artifacts in this
cluster are complete akes that ret one another (see Figure
3B). The only silcrete core in the cluster was assigned to
the MSA based on its paern of reduction, suggesting per-
sistence of MSA technology as late as 30 ka in the region.
Interestingly, the akes have a thin red ‘skin’ that is not
cortical and a bright yellow interior. The color and exterior
surface topography suggest that the rock was heated before
aking, while the bright yellow interior suggests that it was
not heated, potentially highlighting some complexities in
the identication of heat treatment in the area.
Nearby, at the far western edge of the locality, and ap-
parently exposed on the cobble bench by erosion of the ~30
ka sediment unit, is a cluster of 12 preferential and recur-
rent Levallois cores. Eight of these are made from quarite
and two each from silcrete and hornfels. Abundant associ-
ated aking debris suggests signicant potential for rets
but this was not aempted. Similar cores are also abundant
in and around the exposures with major post-Howiesons
Poort and Late MSA components, potentially suggesting
that these relate either to another unknown phase within
MIS 3 or to the Early MSA. The nearby OSL age provides
a minimum age for this assemblage which, combined with
the relative freshness of the artifacts, seems more support-
ive of a Late MSA association.
UITSPANKRAAL 1 (UPK1)
UPK1 is a very large sediment body, however, our work
here was constrained to an area of 96,699m2 due to the pres-
ence of tilled elds along the northern edge. The stack is
14–31m above the Doring River channel and divisible into
four areas. At the western edge are slightly indurated yel-
low sands overlain by active modern dunes. In the central
area are two erosional embayments exposing indurated
red sediments potentially comparable to those in the main
MSA area at UPK7. Along the northern edge is a well-de-
veloped nodular calcrete, probably analogous to that at
UPK9, which forms the ridge on which the tilled eld oc-
curs. Aempts to obtain U/Th ages on these calcretes were
unsuccessful due to excessive detrital content (S910414, see
SOM). Due to the tilled eld, only a small corridor along
the southern edge of the ridge was suitable for survey. At
the eastern edge of the locality is an extensive sheet of par-
tially indurated brown sediment which may be colluvial
rather than aeolian in origin and which is underlain by col-
luvial rocks and gravels.
Our Phase I surveys recorded 1252 artifacts at UPK1
(Figure 4). Unlike the other localities, ESA-assigned arti-
facts—represented entirely by handaxes—are reasonably
well represented, though their distribution is largely con-
strained to the calcrete on the northern edge of the sur-
veyed area. MSA artifacts are distributed across the locality,
though concentrated on the exposed areas of indurated red
sediment. LSA artifacts and poery fragments are largely
restricted to the western edge of the locality, occurring in
blowouts where the slightly consolidated yellow sands are
414 • PaleoAnthropology 2019
183 bifacial pieces that we condently assigned to the Still
Bay (see Figure 5A), as well as four unifacial points and
13 end scrapers. A further 100 bifacially worked pieces are
present in the scaer that are likely also associated with the
Still Bay, though in most cases these were in early stages of
manufacture and could not be condently assigned. Thin-
ning akes associated with the production of points are
abundant across the surface. All of this material appears
to be eroding out from under the low mound on which the
Robberg cluster sits. Phase II analysis was conducted on
this Still Bay cluster and sediment samples for OSL dating
were taken to constrain its age; these results will be pre-
sented elsewhere. In addition to the aked stone artifacts,
a single piece of engraved ochre was located toward the
center of the scaer (see Figure 5B).
DORINGBOS 8 (DB8)
Doringbos 8 sits at the mouth of a short tributary of the
Doring River and comprises a sediment stack approxi-
mately ten meters high, traces of which extend from 3–21m
above the main channel of the Doring. Unlike other stacks,
DB8 has been cut by periodic water ows from the trib-
a series of small mounds across the study area. Capping
this sequence are large swathes of modern vegetated sands,
which are predominantly free of surface artifacts and likely
preserve the sequence of indurated, possibly artifact-bear-
ing, deposits below.
In total, 6747 artifacts were mapped at KH1 during
Phase 1. As with the previously discussed localities, arti-
facts are absent from the modern vegetated upper sands
but common in the erosional areas of the deposit. MSA
and, to a lesser extent, ESA artifacts are common through-
out the lag deposit on the western side of the locality, while
some Pleistocene LSA (Early LSA and Robberg) artifacts
occur on the eastern side. Late MSA cores are abundant
and Post-Howiesons Poort artifacts reasonably common,
but both are diusely scaered among the colluvium; a
few artifacts aributed to the Howiesons Poort and Early
MSA were also recorded in this area. There is a reasonably
well-dened—but small—cluster of Robberg cores located
in a small depression at the top of a low rise on the eastern
side of the locality. Immediately below this in an erosion-
al feature near the base of the sedimentary sequence that
contains a cluster of Still Bay artifacts. The cluster includes
Figure 4. Distribution of artifacts and sediment bodies across Uitspankraal 1 (UPK1). (A) Dense cluster of recurrent Levallois cores,
akes and blades made from local blue-grey quarite.
Doring River Archaeology Project • 415
Figure 5. Distribution of artifacts and sediment bodies across Klein Hoek 1 (KH1). (A) A selection of bifacial pieces from the Still Bay cluster. (B) Engraved ochre in proximity
to Robberg and Wilton artifacts.
416 • PaleoAnthropology 2019
at surfaces—one to the east and one to the west—the for-
mer of which is about 1.5m higher than the laer. The en-
tire sediment stack is 6–14m above the current channel of
the Doring River. Like DB8, PL1 was originally interpreted
as a slackwater accumulation, and like DB8 it has been cut
by activation of the tributary though it does not preserve
an intact section as at DB8. OSL determinations in the 2010
excavation of the eastern (higher) mound surface at PL1 re-
turned ages of 60.8±5.2 and 58.8±5.3 ka from 0.8m and 1.5m
below surface respectively (Mackay et al. 2014b).
The entire surface assemblage at PL1 was originally as-
signed to the Late MSA based on both its distinctive char-
acteristics and the OSL ages, combined with the absence
of indicators from MIS 4. The results of our more compre-
hensive surface survey are broadly in line with that initial
observation, though produce some valuable new observa-
tions. A total of 636 artifacts was recorded at PL1, with 193
assigned to the MSA; only four artifacts were assigned to
the LSA. Of the other artifacts that could be assigned to in-
dustries, the vast majority belonged to the Late MSA. The
four LSA-assigned artifacts were all allocated to the Early
LSA. Surprisingly, the locality produced no clear evidence
for occupation after ~22 ka, which is the local start of the
Robberg. A small number of post-Howiesons Poort arti-
facts was observed, however, consistent with a formation
age for the upper surface of around 58–61 ka. Both of these
post-Howiesons Poort artifacts occurred on the northern
edge of the lower mound surface. This area was also no-
tably richer in silcrete (20 out of 480, 4.2%) than the upper
mound surface (3 out of 157 pieces, 1.9%). Given that the
lower mound surface is ~1.5m below the upper mound sur-
face, and that the OSL ages were recovered from 0.8–1.5m
below the surface of that mound, it may be that the lower
surface eectively dates to ~58–61 ka, and thus formed
within the post-Howiesons Poort interval.
DISCUSSION
The Doring River corridor was occupied from at least the
Middle Pleistocene, and heavily occupied from the MSA
into the historic period. Though we only recorded cores
and implements—which typically account for quite small
proportions of assemblages in the region—sample sizes
were robust on all of the sediment stacks that we have so
far studied. Visible artifact density was likely inuenced
by surface erosion; badly denuded localities like KH1 (0.35
cores and implements/m2), UPK9 (0.32/m2) and PL1 (0.22/
m2) have higher densities of artifacts than those such as
DB8 (0.06/m2) and UPK1 (0.01/m2) on which both vegeta-
tion and the recent dunes have been preserved. Only in
the western and northern parts of UPK7 do we see clear
evidence for denuded surfaces that have low densities of
artifacts. Interestingly much of this surface appears to date
to around 30 ka, hypothetically allowing for accumulations
of LSA material; these instead are concentrated towards the
crest of the stack.
That the Doring River is a major source of stone for
artifact manufacture likely increased the abundance of ar-
tifacts, particularly cores, along much of its course. How-
utary, dividing it into separate northern (15,320m2) and
southern lobes (14,218m2). The north face of the southern
lobe provides a six-and-a-half-meter section with visible
laminations (though no visible lenses of artifacts). We cur-
rently interpret this stack as a slackwater deposit, result-
ing from backooding of the tributary by the Doring River.
Erosion is common across the surfaces of both the southern
and northern lobes, with large blowouts occurring in both.
A minor drainage channel has formed at the junction of
the northern lobe and the underlying bedrock on the north
side, further accelerating sediment loss.
Steep scree slopes characterize the margins of the small
tributary. Surface sediments across the two lobes are rela-
tively uniform, consisting of compact, light brown ne
sands. Although these exposed sediments are likely to vary
in age, their uniformity aligns with available subsurface
data. The 6.5m sedimentary sequence exposed between the
two lobes by stream down-cuing is characterized by the
cyclical deposition of ne sands that ne upward. In the
north-west portion of the northern study lobe, more inten-
sive erosion has exposed a colluvial surface. Slightly to the
east, outside the study area and near the upper reaches of
the short tributary, small patches of a brownish-red paleo-
sol have been preserved, as have similarly small and isolat-
ed traces of more slackwater deposit. The paleosol is likely
older than the slackwater deposit, although it may repre-
sent a soil associated with a particular time period during
slackwater deposition in the tributary that was more con-
ducive to soil formation processes at the upper reaches—
the location where ood water would have been shallower
with a slower rate of deposition.
Phase I survey documented 1814 artifacts at DB8 across
both the northern and southern sections (Figure 6). LSA ar-
tifacts occur at the top of both stacks, with MSA artifacts
below. Quar-dominated LSA artifacts are common in
the uppermost scaers on both lobes, as is chert. On the
northern lobe, the MSA scaer exhibits a central cluster of
silcrete, while being dominated by hornfels and quarite
elsewhere. We expect that the quar-dominated upper
scaers on both lobes are related to Late Holocene occupa-
tion. On the northern lobe there is a clear Wilton cluster
siing just above limited Oakhurst and Robberg signals.
The MSA scaers on both lobes include post-Howie-
sons Poort, Howiesons Poort, and Still Bay components,
though these do not occur in the expected sequence relative
to elevation in the either lobe. This may reect either redis-
tribution during erosion of the blowout, a complex depo-
sitional/sedimentation sequence (unlikely given the visible
laminations), or mis-assignment of artifacts to industries.
Only excavation could resolve these possibilities.
PUTSLAAGTE 1 (PL1)
Putslaagte 1 is the only previously excavated and pub-
lished locality that we have so far surveyed. The locality
comprises a single low mound covering 2941m2 (Figure 7)
situated on a distal spur at the conuence of the Putslaagte
and the Doring Rivers. We did not distinguish multiple
sediment bodies at PL1, though the mound includes two
Doring River Archaeology Project • 417
Our data do reveal interesting paerns of clustering—
which we loosely dene here as spatially coherent distri-
butions of similar artifacts—at dierent spatial extents. At
the nest scale, the Doring River sediment stacks appear to
preserve occasional evidence of clustering that we might
consider representative of ‘events.’ We see this most nota-
bly in the opportunistic ret sets at UPK7 and UPK1. In
total, during the 2019 season we identied nine ret sets, to
go with the three ret sets identied during our previous
work (Low et al. 2017). All of the 2019 examples occurred
in low density areas suggesting that rets are probably
quite prevalent but often dicult to discern due to both
the abundance of archaeology and the fact that our current
surveys are focussed only on cores and implements. That
these ret sets extend into the MSA supports the inferred
ever, it is possible that the abundance of material on the
sediment stacks reects similar occupational decisions to
those inferred by Sampson (1984) in his extensive Karoo
surveys where locations selected for occupation were typi-
cally sandy, free of rock, and located close to—but not im-
mediately on—reliable waterholes. All of the localities we
examined were located at least ~100–150m from the current
channel of the Doring River. Our ability to infer paerns of
occupation at the landform scale, though, is limited by our
survey strategy. While we surveyed a buer of 5m around
each stack, we have yet to conduct any systematic o-
site surveys in the area, and thus whether these sediment
stacks were more heavily occupied than other landforms
in the Doring catchment is something of which we cannot
yet be sure.
Figure 6. Distribution of artifacts and sediment bodies across Doringbos 8 (DB8).
418 • PaleoAnthropology 2019
In addition to spatial clustering, there is evidence for
vertical stacking of sediment units that is likely to warrant
excavation. This is most clearly the case at DB8 where prob-
able Late Holocene material overlies a succession of Wil-
ton, Oakhurst, and Robberg, and then further below where
MSA artifacts from the post-Howiesons Poort, Howiesons
Poort, and Still Bay were mapped. Dating of the exposed
section at DB8 is currently underway, though on the ba-
sis of the available evidence it seems plausible that a long
cultural sequence is present at this locality. Vertical stack-
ing of Late Holocene, Wilton, Oakhurst, and Late MSA also
seems to occur in the northern erosional face of UPK7.
At the scale of industries or technocomplexes, clus-
tering does seem to take dierent forms. For example,
the Still Bay occurs as major (e.g., KH1) and minor (e.g.,
UPK7) clusters which are in both cases tightly spatially
constrained. This is also true for the post-Howiesons Poort,
Early LSA, Oakhurst, Wilton, and possibly the Robberg. So
far, however, Late MSA clusters are always distributed as
smears over relatively large areas. This may reect the ex-
tended time interval that the Late MSA represents, which at
~25 kyr is around 3–5 times longer than the duration of the
other industries. This is not to say that the dening features
we have provided for the Late MSA hold for that entire du-
ration. Indeed, this degree of technological stability would
spatial integrity of the deposits across at least the last 30
kyr, though putative Still Bay at KH1 suggests that spatial
integrity extends much deeper into the MSA.
At a somewhat larger scale—which we might term
‘aggregates’—we found clusters with numbers of artifacts
suggestive of either repeated occupation of specic loca-
tions on stacks or occupation by large numbers of people
within limited areas. This occurs whether the aggregate
is represented by retouched implements (e.g., Still Bay at
KH1, Oakhurst and ‘Woodlot/duckbill scrapers’ at UPK9)
or cores (e.g., post-Howiesons Poort and Early LSA at
UPK7, Robberg at UPK9 and KH1). Some of these aggre-
gates are unquestionably constrained by exposure visibil-
ity of appropriately-aged sediment surfaces. This is most
notably true of the Still Bay at KH1, which likely extends
under the adjacent mound, and the Early LSA at UPK7,
which is entirely co-extensive with the blowout in which
it is observed. It may also be true of the Robberg at UPK9
that likely continues across much of the indurated yellow
sediment unit which is currently covered by recent dunes
and vegetation. There is clearly signicant excavation po-
tential in these areas. In other cases, such as the eastern
Oakhurst cluster at UPK9, the extent of the aggregate is not
constrained by visibility and likely reects the full spatial
distribution of the phenomenon.
Figure 7. Distribution of artifacts and sediment bodies across Putslaagte 1 (PL1).
Doring River Archaeology Project • 419
facts between dierent industries. The Late Holocene and
Wilton are common in open sites despite having been rare
in rock shelter excavations in the catchment. This is even
more acutely true for the Late MSA. The Howiesons Poort,
on the other hand, shows an inverted paern of distribu-
tion—common in shelters, rare in the open. Only the Rob-
berg and to a lesser extent the Oakhurst are common in all
contexts. Conversely, the Early LSA is rare throughout.
These paerns, though, presume that we can identify
artifacts from given periods using similar identifying char-
acteristics in all contexts. We know from our past work in
the catchment that this approach is fraught with issues.
Technological systems are responsive to the availability of
raw materials at a minimum, and probably also of water,
as well as other variables with spatial and temporal dis-
tributions that are less easy to identify. We are also con-
fronted with coherent clusters of material that we currently
cannot place within the local sequence at all, though they
may be expressions of entities known from the broader re-
gion. Both of these observations converge on the same solu-
tion, however—excavations are required to move beyond
the process of dating-by-inference on which we currently
rely. Survey work away from the sediment bodies is a fur-
ther necessity before our broader project aims can be met,
alongside completion of analysis of artifact assemblages
and dating of sediments from the rock shelters we have ex-
cavated. The integration of these data sets should allow our
broader goals to be met.
ACKNOWLEDGEMENTS
The Doring River Archaeology Project has been sup-
ported by grants from the Australian Research Council
(DE130100068 & FT160100139) to AM. CA is supported
by a post-doctoral fellowship from the Faculty of Science,
Medicine and Health at the University of Wollongong.
ENDNOTES
1We use the term ‘sediment stacks’ to refer to large accumulations of sedi-
ment along the Doring River as described in the Identication of Ar-
chaeological Localities section. The term ‘locality’ is used to refer to
any concentration of archaeological material, whether on a sediment
stack or not. So far, all of the sediment stacks we have identied have
also been localities in this sense.
REFERENCES
Andrefsky, W.A.J. 1994. Raw material availability and the
organization of technology. American Antiquity 59(1):
21–34.
Bailey, G. 2007. Time perspectives, palimpsests and the ar-
chaeology of time. Journal of Anthropological Archaeology
26(2): 198–223.
Bamforth, D.B. 1991. Technological organisation and hunt-
er-gatherer land use: a California example. American
Antiquity 56(2): 216–234.
Barham, L.S. 1989. The Later Stone Age of Swaziland. Ph.D.
thesis, Philadelphia, University of Pensylvania.
Barton, C.M. and J. Riel-Salvatore. 2014. The formation of
lithic assemblages. Journal of Archaeological Science 46:
334–352.
be unusual against the broader paern of the last 75 kyr.
Alternatively, then, if the simple prepared core systems on
at pebbles that we associate with the Late MSA are char-
acteristic of only a portion of the period 50–25 ka, then the
diuse clustering we appear to see in the Late MSA may
have a behavioral explanation. While we cannot evaluate
this prospect with our available data, as noted below, the
broader paern of Late MSA artifact distribution is quite
dierent to that of the other industries we have considered.
Consistent with previous work (Hallinan and Park-
ington 2017; Mackay et al. 2018), the Howiesons Poort is
extremely poorly represented in our data so far. This is in
spite of the fact that the Howiesons Poort—in contrast to
most other industries—has clear identifying characteristics
for implements, cores, and akes. It is usually even possible
to identify it through the abundance of silcrete in combi-
nation with the above markers. The paucity of Howiesons
Poort artifacts in our surveys also stands in stark contrast to
their abundance in excavated samples in the Doring catch-
ment and across the Winter Rainfall Zone more broadly
(Mackay et al. 2014a). The outcome here is essentially the
inverse of that noted for the Late MSA, which was well-rep-
resented in our surveys but rare in shelters. Though fasci-
nating, our sediment stack surveys remain incomplete, and
as we have noted above do not take in open-air locations
away from the river. The landscape-scale distribution of the
Howiesons Poort in the Doring River valley remains only
partially sketched thus far.
One other industry, or set of industries depending on
the classication system (e.g., Lombard et al. 2012; Mackay
et al. 2014a), that is poorly represented for dierent reasons
is the Early MSA. We consider it likely that many Early
MSA artifacts occur within our recorded sample, yet we
currently lack condent identifying characteristics for this
period. The only retouched ake types that we know to oc-
cur regularly in regional expressions of the Early MSA—
notched and denticulate akes—are also known to occur in
the Late MSA and Early LSA (Porraz et al. 2016; Will et al.
2015). Of the denticulates that we recorded, most occurred
as isolated instances, some of which were associated with
Late MSA cores. So far, no clusters of denticulate pieces
have been observed. We have also yet to nd any cores
capable of producing the kinds of very large blades and
akes often associated with the Early MSA, though blades
>80mm were observed with recurrent Levallois cores at
UPK1. Otherwise, this industry remains beyond the reach
of our current approach.
CONCLUSIONS
The long-term objective of the Doring River Archaeology
Project is to understand the organization of technology at
the landscape scale through at least the Late Pleistocene.
The localities that we have worked on so far provide an
encouraging wealth of material from this interval, and dis-
play a spatial structure indicative of signicant horizontal
and occasionally vertical integrity in spite of extensive and
likely recent erosion. Preliminary paerns suggest strik-
ing dierences in the abundance and distribution of arti-
420 • PaleoAnthropology 2019
Acheulian and its transition to the Middle Stone Age
in southern Africa: the question of the Fauresmith. In-
ternational Journal of Evolutionary Biology 2011: 961401.
Holdaway, S. and M. Douglass. 2015. Use beyond manu-
facture: non-int stone artifacts from Fowlers Gap,
Australia. Lithic Technology 40(2): 94–111,
Howland, M.D., I.W.N. Jones, M. Najjar, and T.E. Levy
2018. Quantifying the eects of erosion on archaeologi-
cal sites with low-altitude aerial photography, struc-
ture from motion, and GIS: a case study from southern
Jordan.” Journal of Archaeological Science 90: 62–70.
Jarvis, A., H.I. Reuter, A. Nelson, and E. Guevara. 2008.
Hole-lled SRTM for the globe Version 4, available
from the CGIAR-CSI SRTM 90m Database (hp://srtm.
csi.cgiar.org).
Jerardino, A. and R. Yates. 1996. Preliminary results from
excavations at Steenbokfontein Cave: implications for
past and future research. South African Archaeological
Bulletin 51: 7–16.
Kaplan, J. 1990. The Umhlatuzana Rock Shelter seqeunce:
100 000 years of Stone Age history. Natal Museum Jour-
nal of Humanities 2: 1–94.
Kelly, R.L. 1995. The Foraging Spectrum: Diversity in Hunter-
Gatherer Lifeways. Washington, D.C., Smithsonian Insti-
tution Press.
Lin, S.C., M.J. Douglass, and A. Mackay. 2016. Interpreting
MIS 3 artefact transport paerns in southern Africa us-
ing cortex ratios: an example from the Putslaagte val-
ley, Western Cape. South African Archaeological Bulletin
71(204): 173–180.
Loftus, E., J. Sealy, and J. Lee-Thorp. 2016. New radiocar-
bon dates and Bayesian models for Nelson Bay Cave
and Byneskranskop 1: implications for the South Af-
rican Later Stone Age sequence. Radiocarbon 58(02):
365–381.
Lombard, M., L. Wadley, J. Deacon, S. Wurz, I. Parsons, M.
Mohapi, J. Swart, and P. Mitchell. 2012. South African
and Lesotho Stone Age sequence updated. South Afri-
can Archaeological Bulletin 67(195): 120–144.
Low, M., A. Mackay, and N. Phillips. 2017. Understanding
Early Later Stone Age technology at a landscape-scale:
evidence from the open-air locality Uitspankraal 7
(UPK7) in the Western Cape, South Africa. Azania: Ar-
chaeological Research in Africa 52(3): 373–406.
Low, M.A. and A. Mackay. 2016. The late Pleistocene mi-
crolithic at Putslaagte 8 rockshelter in the Western
Cape, South Africa. South African Archaeological Bulletin
71(204): 146–159.
Low, M.A. and A. Mackay. 2018. The organisation of Late
Pleistocene Robberg blade technology in the Doring
River Catchment, South Africa. Journal of African Ar-
chaeology 16: 168–192.
Mackay, A., C. Cartwright, S. Heinrich, M. Low, M.
Stahlschmidt, and T.E. Steele. in press. Excavations at
Klipfonteinrand reveal local and regional paerns of
adaptation and interaction through MIS 2 in southern
Africa. Journal of Palaeolithic Archaeology.
Mackay, A., E. Hallinan, and T.E. Steele. 2018. Provision-
Beaumont, P.B. 1978. Border Cave. M.A. thesis, Cape Town,
University of Cape Town.
Binford, L.R. 2001. Constructing Frames of Reference: An Ana-
lytical Method for Archaeological Theory Building Using
Ethnographic and Environmental Data Sets. Berkeley and
Los Angeles, University of California Press.
Bleed, P., M.J. Douglass, A. Sumner, M. Behrendt, and A.
Mackay. 2017. Photogrammetrical assessment of pro-
cedural paerns and sequential structure in “handaxe”
manufacture: a case study along the Doring River of
South Africa. Lithic Technology 2017: 1–11.
Bousman, C.B. and J.S. Brink 2017. The emergence, spread,
and termination of the Early Later Stone Age event in
South Africa and southern Namibia. Quaternary Inter-
national 495: 116–135.
Carter, P.L. 1978. The Prehistory of Eastern Lesotho. Ph.D. the-
sis, Cambridge, UK, University of Cambridge.
Carter, P.L., P.J. Mitchell, and P. Vinnicombe. 1988. Sehong-
hong: The Middle and Later Stone Age Industrial Sequence
at a Lesotho Rockshelter. Oxford, British Archaeological
Reports International Series 406. Archaeopress.
Deacon, H.J. 1976. Where Hunters Gathered: A Study of Ho-
locene Stone Age People in the Eastern Cape. Claremont,
South African Archaeological Society.
Deacon, H.J. 1979. Excavations at Boomplaas Cave - A se-
quence through the Upper Pleistocene and Holocene in
South Africa. World Archaeology 10(3): 241–257.
Deacon, J. 1984. The Later Stone Age in Southernmost Africa.
Oxford, Archaeopress.
Feilden, H.W. 1884. Notes on Stone Implements from South
Africa. Journal of the Anthropological Institute of Great
Britain and Ireland 13: 162–174.
Fisher, E., R.-M. Albert, G.A. Botha, H. Cawthra, I. Esteban,
J.W. Harris, Z. Jacobs, A. Jerardino, C.W. Marean, F.H.
Neumann, J. Pargeter, M. Poupart, and J. Venter. 2013.
Archaeological reconnaissance for Middle Stone Age
sites along the Pondoland Coast, South Africa. Paleo-
Anthropology 2013: 104–137.
Gooch, W.D. 1882. The Stone Age of South Africa. Journal of
the Royal Anthropological Institute 11: 124–183.
Goodwin, A.J.H. and C. van Riet Lowe. 1929. The Stone Age
Cultures of South Africa. Edinburgh, Neill and Co.
Hall, S. 1990. Hunter-Gatherer-Fishers of the Fish River Basin:
A Contribution to the Holocene Prehistory of the Eastern
Cape. D.Phil. thesis, Stellenbosch, University of Stellen-
bosch.
Hallinan, E. and J. Parkington. 2017. Stone Age landscape
use in the Olifants River Valley, Clanwilliam, Western
Cape, South Africa. Azania: Archaeological Research in
Africa 52(3): 324–372.
Harmand, S., J.E. Lewis, C.S. Feibel, C.J. Lepre, S. Prat, A.
Lenoble, X. Boes, R.L. Quinn, M. Brenet, A. Arroyo, N.
Taylor, S. Clement, G. Daver, J.P. Brugal, L. Leakey,
R.A. Mortlock, J.D. Wright, S. Lokorodi, C. Kirwa, D.V.
Kent, and H. Roche. 2015. 3.3-million-year-old stone
tools from Lomekwi 3, West Turkana, Kenya. Nature
521(7552): 310–315.
Herries, A.I. 2011. A chronological perspective on the
Doring River Archaeology Project • 421
ology of Prehistoric Coastlines, G.N. Bailey and, J.E. Park-
ington (eds.). Cambridge, Cambridge University Press,
pp. 22–41.
Paxton, B. 2008. The Inuence of Hydraulics, Hydrology and
Temperature on the Distribution and Habitat Use and Re-
cruitment of Threatened Cyprinids in a Western Cape River,
South Africa. Ph.D. thesis, Cape Town, University of
Cape Town.
Phillips, N. in prep. Out in the Open: A Geoarchaeological Ap-
proach to Open-air Surface Archaeology in the Semi-arid
Interior of South Africa’s Western Cape. Ph.D. thesis, Wol-
longong, University of Wollongong.
Phillips, N., J. Pargeter, M. Low and A. Mackay. 2018.
Open-air preservation of miniaturised lithics: experi-
mental research in the Cederberg Mountains, south-
ern Africa. Archaeological and Anthropological Sciences
11(11): 5851–5877.
Plaske, J. 2012. The Stone Artefacts of Putslaagte 8: Explor-
ing Variation in Stone Artefact Assemblages of the Western
Cape. Unpublished B.A. Honours, Cape Town, Univer-
sity of Cape Town.
Porraz, G., V. Schmid, C. Miller, C. Tribolo, C. Cartwright,
A. Charrié-Duhaut, M. Igreja, S.M. Mener, N. Mer-
cier, P. Schmid, N.J. Conard, J.P. Texier, and J. Parking-
ton. 2016. Update on the 2011 excavation at Elands Bay
Cave (South Africa) and the Verlorenvlei Stone Age.
Southern African Humanities 29: 33–68.
Porraz, G., P.-J. Texier, W. Archer, M. Piboule, J.-P. Rigaud,
and C. Tribolo. 2013. Technological successions in the
Middle Stone Age sequence of Diepkloof Rock Shelter,
Western Cape, South Africa. Journal of Archaeological
Science 40(9): 3376–3400.
Quick, L.J. and F.D. Eckardt. 2015. The Cederberg: a rug-
ged sandstone topography. In Landscapes and Landforms
of South Africa, S. Grab and J. Knight (eds.), Dordrecht,
Springer, pp. 85-93.
Sadr, K. 2015. Livestock rst reached southern Africa in
two separate events. PLoS One 10(8): e0134215
Sampson, C.G. 1968. The Middle Stone Age Industries of the
Orange River Scheme Area. Bloemfontein, National Mu-
seum.
Sampson, C.G. 1984. Site clusters in the Smitheld sele-
ment paern. South African Archaeological Bulletin
39(139): 5–53.
Shea, J.J. 2014. Sink the Mousterian? Named stone tool in-
dustries (NASTIES) as obstacles to investigating hom-
inin evolutionary relationships in the Later Middle Pa-
leolithic Levant. Quaternary International 350: 169–179.
Singer, R. and J. Wymer. 1982. The Middle Stone Age at Kla-
sies River Mouth in South Africa. Chicago, University of
Chicago Press.
Smith, A.B. and M.R. Ripp. 1978. An archaeological recon-
naissance of the Doorn/Tanqua Karoo. South African Ar-
chaeological Bulletin 33: 118–133.
Thackeray, A.I. 1977. Stone Artefacts from Klipfonteinrand.
B.A. Honours, Cape Town, University of Cape Town.
Villa, P., S. Soriano, T. Tsanova, I. Degano, T. Higham, F.
d’Errico, L. Backwell, J. Lucejko, M.P. Colombini, and
ing responses to environmental variation in the late
Pleistocene of southern Africa, MIS 5–2. In Lithic Tech-
nological Organization and Paleoenvironmental Change. E.
Robinson and F. Sellet (eds.), Dordrecht, Springer, pp.
13–36.
Mackay, A., Z. Jacobs, and T.E. Steele. 2015. Pleistocene ar-
chaeology and chronology of Putslaagte 8 (PL8) rock-
shelter, Western Cape, South Africa. Journal of African
Archaeology 13(1): 71–98.
Mackay, A., B.A. Stewart, and B.M. Chase. 2014a. Coales-
cence and fragmentation in the late Pleistocene archae-
ology of southernmost Africa. Journal of Human Evolu-
tion 72: 26–51.
Mackay, A., A. Sumner, Z. Jacobs, B. Marwick, K. Blu, and
M. Shaw. 2014b Putslaagte 1 (PL1), the Doring River,
and the later Middle Stone Age in southern Africa’s
Winter Rainfall Zone. Quaternary International 350: 43–
58.
Magnani, M., M. Douglass, and S.T. Porter. 2016. Closing
the seams: resolving frequently encountered issues in
photogrammetric modelling. Antiquity 90(354): 1654–
1669.
Mazel, A.D. 1989. People making history: the last ten thou-
sand years of hunter-gatherer communities in the
Thukela Basin. Natal Museum Journal of Humanities 1:
1–168.
McPherron, S.P., Z. Alemseged, C.W. Marean, J.G. Wynn,
D. Reed, D. Geraads, R. Bobe, and H.A. Bearat. 2010.
Evidence for stone-tool-assisted consumption of ani-
mal tissues before 3.39 million years ago at Dikika,
Ethiopia. Nature 466(7308): 857–860.
Mitchell, L.J. 2009. Belongings: Property, Family, and Identity
in Colonial South Africa (an exploration of frontiers, 1725-c.
1830). New York, Columbia University Press.
Mitchell, P.J. 2000. The organization of Later Stone Age lith-
ic technology in the Caledon Valley, Southern Africa.
African Archaeological Review 17(3): 141–176.
Mitchell, P.J. 2008. Developing the archaeology of Marine
Isotope Stage 3. South African Archaeological Society
Goodwin Series 10: 52–65.
Nackerdien, R. 1989. Klipfonteinrand 2: A Sign of the Times.
B.A. Honours, Cape Town, University of Cape Town.
Nelson, M. 1991. The study of technological organization.
Archaeological Method and Theory 3: 57–100.
Neumark, S.D. 1957. Economic Inuences on the South African
Frontier 1652-1836. Stanford, USA, Stanford University
Press.
Opperman, H. 1996. Strathalan Cave B, north-eastern Cape
Province, South Africa: evidence for human behaviour
29,000-26,000 years ago. Quaternary International 33:
45–53.
Parkington, J.E. 1980. The Elands Bay cave sequence: cul-
tural stratigraphy and subsistence strategies. Proceed-
ings of the Eighth Pan-African Congress of Prehistory and
Quaternary Studies, Nairobi, pp. 315–320.
Parkington, J.E., Poggenpoel, C.A., Buchanan, W., Robery,
T., Manhire, A.H., and Sealy, J. 1988. Holocene coastal
selement paerns in the Western Cape. In The Archae-
422 • PaleoAnthropology 2019
to the excavations, stratigraphy and dating. Southern
African Humanities 18: 1–26.
Wendt, W.E. 1972. Premilinary report on an archaeological
research programme in South West Africa. Cimbebasia
B 2: 1–61.
Wilkins, J. and M. Chazan. 2012. Blade production ∼500
thousand years ago at Kathu Pan 1, South Africa: sup-
port for a multiple origins hypothesis for early Middle
Pleistocene blade technologies. Journal of Archaeological
Science 39(6): 1883–1900.
Will, M., A. Mackay, and N. Phillips. 2015. Implications of
Nubian-like core reduction systems in Southern Africa
for the identication of early modern human disper-
sals. PLoS One 10(6): e0131824.
P. Beaumont. 2012. Border Cave and the beginning of
the Later Stone Age in South Africa. Proceedings of the
National Academy of Sciences USA 109(33): 13208–13213.
Visser, H.N. and J.N. Theron. 1973. 3218 Clanwilliam. Geo-
logical Series. Pretoria, Government Printer, South Af-
rica.
Volman, T.P. 1981. The Middle Stone Age in the Southern Cape.
Ph.D. thesis, Chicago, University of Chicago.
Wadley, L. 1984. Later Stone Age Hunter-Gatherers of the
Southern Transvaal: Social and Ecological Interpretation.
Oxford, British Archaeological Reports International
Series 380.
Wadley, L. 1997. Rose Coage Cave: archaeological work
1987 to 1997. South African Journal of Science 93: 439–444.
Wadley, L. and Z. Jacobs. 2006. Sibudu Cave: background
