Content uploaded by Genevieve Dewar
Author content
All content in this area was uploaded by Genevieve Dewar on Oct 11, 2017
Content may be subject to copyright.
Stable isotope evidence of late MIS 3 to middle
Holocene palaeoenvironments from Sehonghong
Rockshelter, eastern Lesotho
EMMA LOFTUS,
1
* BRIAN A. STEWART,
2
GENEVIEVE DEWAR
3
and JULIA LEE-THORP
1
1
Research Laboratory for Archaeology and the History of Art, University of Oxford, Oxford, UK
2
Museum of Anthropological Archaeology, University of Michigan, USA
3
Department of Anthropology, University of Toronto (Scarborough), Canada
Received 21 June 2015; Revised 4 September 2015; Accepted 5 October 2015
ABSTRACT: Evidence for human occupation of southern Africa’s high-altitude Maloti–Drakensberg Mountains is
surprisingly common in the last glacial, yet the attraction of this relatively severe, cold region for hunter-foragers
remains unclear. Sehonghong Rockshelter (1870 m asl), in the eastern Lesotho Highlands, provides evidence for
human occupation spanning Marine Isotope Stage 3 through the late Holocene. Excellent organic preservation
provides opportunities for establishing multiple palaeoenvironmental proxy records to address this conundrum. In
high-altitude zones, the proportions of C
3
and C
4
plants archived in soil organic matter and faunal enamel provide
sensitive indicators of past temperature shifts. We first extended the radiocarbon chronology to ca. 35 ka using
ABOx-SC radiocarbon dates of charcoals. Next we analysed stable isotopes in soil organic matter from the
sedimentary sequence, and in faunal tooth enamel from the newly dated lower strata. The results suggest,
predictably, that C
3
vegetation and low temperatures prevailed until early warming at ca. 15 ka, with a series of
sharp shifts thereafter. Low values for d
13
C and d
18
O in faunal enamel ca. 33 ka suggest a negative temperature
excursion at this time, and potentially greater precipitation as snowfall in the highlands compared with lower
altitudes. Copyright #2015 John Wiley & Sons, Ltd.
KEYWORDS: Lesotho Highlands; Middle and Later Stone Age; palaeoenvironments; radiocarbon dates; stable isotopes.
Introduction
The Maloti–Drakensberg Mountains of Lesotho are a chal-
lenging environment even today under relatively mild Holo-
cene conditions, yet pulsed human occupations are evident
at several sites in the highlands since at least Marine Isotope
Stage (MIS) 5 (Stewart et al., 2012). Persistent, if intermittent,
exploitation of the highlands shows that such habitats
featured strongly in human settlement decisions. Their occu-
pation may reflect a diverse range of push–pull factors that
structure human use of highland landscapes (Stewart et al.,
2012, in press). Interrogation of these decisions requires local
environmental records that attest to the conditions actually
experienced by these groups. Such records, however, are
scarce and insufficient to meet this challenge at present.
A cluster of sites in south-eastern Africa that span a marked
ecological gradient are providing the archaeological and
environmental sequences necessary to examine environmen-
tal context, demographic flux and behavioural flexibility in
detail across the long time spans of the later Middle (MSA)
and Later Stone Age (LSA), and provide context for the
occupational and technological sequence at Sehonghong.
Notably, Melikane, a nearby site at a similar altitude,
contains evidence for a series of pulsed occupations spanning
ca. 80 ka. Stewart et al. (2012, in press) suggest that some
occupation episodes coincide with periods of aridity and
absence of occupation in the interior of southern Africa. They
hypothesize that the more reliable moisture sources at higher
elevations may have drawn people to the uplands, despite
otherwise severely cold conditions. In addition, the region
features several other archaeological sites that span the MSA/
LSA transition, including Rose Cottage Cave and Umhlatu-
zana in South Africa. Many well-studied sequences in South
Africa contain a hiatus between the MSA and LSA, which
makes this an important region for assessing the MSA/LSA
transition in southern Africa. This paper contributes to these
questions by examining temperature-driven shifts in vegeta-
tion changes around Sehonghong from stable isotope analyses
of sedimentary soil organic matter and faunal enamel records
in the site.
Sehonghong
Sehonghong (29˚460S, 28˚470E; Fig. 1) is a large sandstone
rockshelter in the Maloti–Drakensberg Mountains of eastern
Lesotho (Thaba-Tseka District) with a record of human
occupation stretching from the late Holocene into early MIS
3 at ca. 57 ka (Jacobs et al., 2008). The site faces west-north-
west, and is situated on the southern side of the Sehonghong
River, a tributary of the Senqu (Orange) River, at an altitude
of ca. 1870 m asl. Highland Lesotho falls within the Grassland
biome of southern Africa’s summer rainfall zone.
Sehonghong was first excavated in artificial 10 cm spits by
Patrick Carter in 1971 (Carter, 1976; Carter et al., 1988). In
1992, Peter Mitchell opened a new trench in which he
conducted stratigraphic excavations that ceased in levels
thought to be transitional between the LSA and MSA
(Mitchell, 1996c). The underlying MSA levels in Mitchell’s
trench are being re-investigated as part of the ‘Adaptations to
Marginal Environments during the Middle Stone Age’
(AMEMSA) project, which seeks to explore the adaptive range
of early human behaviour in some of southern Africa’s more
challenging habitats. The site’s earlier deposits (of Carter’s
original 1971 trench) were dated by single-grain optically
stimulated luminescence (OSL), which situated the age of the
basal deposits in early MIS 3 (Jacobs et al., 2008). Table 1
shows all previously published radiocarbon and OSL dates
from the site, newly calibrated using the updated SHCal13
calibration curve (Hogg et al., 2013), in sequence with
radiocarbon dates undertaken as part of this study (discussed
below). Sehonghong’s pulsed depositional sequence spans
MSA levels containing post-Howiesons Poort (or ‘Sibudan’ cf.
Correspondence to: E. Loftus, as above
E-mail: emma.loftus@rlaha.ox.ac.uk
Copyright #2015 John Wiley & Sons, Ltd.
JOURNAL OF QUATERNARY SCIENCE (2015) ISSN 0267-8179. DOI: 10.1002/jqs.2817
Lombard et al., 2012) and final MSA assemblages, and LSA
levels from the early Robberg to the post-Classic Wilton with
pottery (e.g. Mitchell, 1995, 1996a,c; Jacobs et al., 2008;
Plug and Mitchell, 2008). The site is notable in the region
for exceptional preservation of organic remains in the late
Pleistocene layers (Mitchell, 1996a), and has a robust Holo-
cene and Terminal Pleistocene chronology for the region
(Carter and Vogel, 1974; Carter, 1976; Carter et al., 1988;
Mitchell and Vogel, 1994).
Environmental Background
Modern climatic and environmental variability in the Lesotho
uplands is strongly determined by topographic variation
within the Maloti–Drakensberg Mountains, the highest in
Africa and part of the Great Escarpment that demarcates the
interior African plateau. Altitudinal temperature gradients
have marked effects on environmental variability within
Lesotho, in contrast to the rest of southern Africa where
vegetation patterning is primarily structured by the amount
and season of precipitation. The highlands receive reliable
rainfall from the uplift of moist Indian Ocean coastal air
masses over the Maloti–Drakensberg during summer, and are
a stable source of runoff for the lower-lying regions in the
rain-shadow cast to the west (Tyson and Preston-Whyte,
2000). South Africa’s largest river, the Orange (known in
Lesotho as the Senqu), begins its journey across the arid
interior plateau in the Lesotho highlands. Thus, the Lesotho
Highlands experience an extreme range of climatic condi-
tions, with occasional heavy snowfall in the uplands during
winter, and mild, wet summers during which rivers swell and
become impassable.
Grassland dominates the Lesotho and probably has done so
throughout the Late Pleistocene and Holocene (Plug, 1997).
Food resources for humans are scarce and patchily distrib-
uted. They can be broadly grouped into three ‘resource
zones’ according to altitude, with the river valleys providing
the greatest range of foraging opportunities, grasslands suit-
able for grazing, dominated by C
4
Themeda triandra (Rooig-
ras), with few trees or shrubs on the plateaux between river
valleys, and finally, less palatable but more cold-tolerant C
3
Festuca species dominating the grassland composition with
increasing altitude (Vogel et al., 1978; Cowling, 1983;
Mitchell, 1996a). In the absence of intense cultivation, the
immediate environment of Sehonghong today would have
been a C
4
Themeda–Cymbopogon–Eragrostis grassland fea-
turing dense stands of C
3
trees and shrubs in the protected
locations of valley bottoms. These locales would have
supplied hunter-gatherers with firewood and edible fruits and
several other edible plants, such as geophytes and reeds
(Mitchell, 1996a). Historically, herds of grazing animals
migrated seasonally across the escarpment, from a winter
Figure 1. Map of southern Africa showing the location of Sehonghong and the distribution of C
4
grass species across the region. The lower
percentages of C
4
grass species in the region around Sehonghong are related to the lower growing season temperatures in the high-elevation
Drakensberg Mountains. Map adapted from Lee-Thorp and Talma (2000), which was in turn based on Vogel et al. (1978).
Copyright #2015 John Wiley & Sons, Ltd. J. Quaternary Sci. , Vol. 9999(9999) 1–12 (2015)
JOURNAL OF QUATERNARY SCIENCE
refuge on the eastern side of the Drakensberg/Maloti Moun-
tains (Von Richter, 1971). At the start of summer, when the
new growth of ‘sourveld’ grasslands in eastern and western
Lesotho was still palatable, the herds moved up into the
eastern Lesotho Highlands, moving further west still in
late summer, towards the ‘sweet’ and mixed grasslands of the
Free State, South Africa, which remain palatable year-round
(Smith et al., 2002). Although not observed, grazers may have
also overwintered in the latter, moving eastwards into Lesotho
during the summer. Productive spawning runs of edible fish
such as Labeo capensis (the Orange River mudfish) and Labeo
barbus aeneus (the smallmouth yellowfish) occur seasonally
in the Senqu, whose confluence with the Sehonghong River
is ca. 3 km downstream from the rockshelter. Abundant fish
remains have been found in Sehonghong’s Holocene layers,
although high densities also occur in some Late Pleistocene
levels, particularly those dated to the Pleistocene–Holocene
transition and the Last Glacial Maximum (LGM) (Plug and
Mitchell, 2008).
Stable isotope variation in a topographically
diverse highland region
The distribution of C
3
and C
4
plants is primarily structured
by growing season temperature (Ehleringer et al., 1997).
C
4
grasses dominate the grassy component where maximum
Table 1. All chronometric ages obtained to date for Sehonghong. New AMS radiocarbon dates presented in this paper (OxA-27689–97) are
shown in sequence with published conventional radiocarbon and OSL dates for Sehonghong from the Mitchell and Carter excavations, with
sample context and the archaeological association.
Calibrated age (cal a BP)
Name Unit Cultural association Date From To
Pta-6084GAP Ceramic Wilton 1240 50 1265 980
Pta-885†IX unit 6 Ceramic Wilton 1400 50 1357 1180
Pta-6063GAP Ceramic Wilton 1710 20 1691 1528
Pta-6154GWA Classic Wilton 5950 70 6931 6548
Q-3174‡IX unit 28 Classic Wilton 6870 60 7818 7571
Pta-6278ALP Later Oakhurst 7290 80 8297 7879
Pta-6280ALP Later Oakhurst 7090 80 8007 7697
Pta-6072ALP Later Oakhurst 7210 80 8173 7839
Pta-6083ALP Later Oakhurst 7010 70 7942 7671
Pta-6368SA Oakhurst 9280 45 10 550 10 255
Pta-6057SA Oakhurst 9740 140 11595 10 588
Pta-6065BARF Robberg 11 090 230 13395 12 559
Pta-6282RF Robberg 12 180 110 14 467 13 731
Q-3175‡IX- unit 39 Robberg 12 250 300 15229 13 472
Q-3176‡IX- unit 42 Robberg 12 200 250 15080 13 495
Pta-6062RBL Robberg 12 410 45 14 735 14 125
Pta-6058CLBRF Robberg 12 470 100 15040 14 137
Q-3173‡IX- unit 48 Robberg 12 800 250 15925 14 184
Pta-884†IX unit 50 Robberg 13 000 140 15948 15 109
Q-3172‡IX unit 52 Robberg 13 200 150 16210 15 304
Pta-6060BAS Robberg 15 700 150 19 298 18 600
Q-1452§IX unit 54 Robberg 17 820 270 22246 20 822
Pta-6281BAS Robberg (early phase) 19 400 200 23853 22 824
Pta-6077BAS Robberg (early phase) 20 200 100 24495 23 952
Pta-789†IX unit 60 Robberg (early phase) 20 900 270 25735 24 430
Pta-918†IX unit 72 Transitional MSA/LSA 19 860 220 24 401 23 316
Pta-919†VII unit 80 Transitional MSA/LSA 20 240 230 25015 23 745
Pta-6059Mos (VII) Transitional MSA/LSA 20 500 230 25 272 24 061
Pta-6271RFS (VI) Transitional MSA/LSA 25 100 300 29 890 28 466
Pta-6268RFS (VI) Transitional MSA/LSA 26 000 430 30 983 29 243
Pta-920†VII/VI/V Mixed associations 28 870 520 33 922 31 576
OxA-27689¶162 MSA 25 330 130 29876 29 036
OxA-27690¶163 MSA 28 650 200 33148 31 917
OxA-27691¶167 MSA 29 120 190 33420 32 765
OxA-27692¶169 MSA 29 170 190 33531 32 966
OxA-27693¶1030 MSA 29 200 200 33630 33 095
OxA-27694¶1031 MSA 28 800 190 33739 33 201
OxA-27695¶1036 MSA 30 910 250 34934 34 226
OxA-27696¶1037 MSA 31 030 250 35105 34 485
OxA-27697¶1111A MSA 30 710 240 35 288 34 585
SEH4 (OSL) IV MSA 3 (Volman) 31 600 1400
SEH3 (OSL) III MSA 3 30 300 3400
SEH2 (OSL) III MSA 3 46 500 2500
SEH1 (OSL) II MSA 3 57 600 2300
Mitchell and Vogel (1994); †Carter and Vogel (1974); ‡Carter et al. (1988); §Carter (1976); ¶This study; Jacobs et al. (2008).
All dates are conventional radiocarbon dates, except SEH1-4, which are single-grain OSL ages from the Carter trench. All are radiocarbon dates of
charcoal, except Q-3175 and Q-3176 on bone. The SHCal13 (Hogg et al., 2013) calibration curve was used; calibrated ages are presented at 95%
confidence interval.
Copyright #2015 John Wiley & Sons, Ltd. J. Quaternary Sci. , Vol. 9999(9999) 1–12 (2015)
SEHONGHONG ROCKSHELTER PALAEOENVIRONMENTS
growing season temperatures are above 25 ˚C and minima do
not fall below 10 ˚C (Teeri and Stowe, 1976; Vogel et al.,
1978; Tieszen et al., 1979), the conditions across most of the
summer rainfall region of southern Africa. In mountainous
zones, however, beyond a threshold the proportions of C
4
grasses in the vegetation decrease steadily with altitude and
decreasing temperatures (Fig. 2). In the past when mean
annual temperatures may have been lower, vegetation belts
therefore shifted to lower elevations. Conversely the threshold
favouring C
4
plants shifted upwards during warmer periods.
The relationship between altitude and vegetation patterning
is also affected by aspect. In Lesotho, the C
4
grassland
extends up to ca. 2700 m asl on north-facing slopes, with
decreasing C
4
elements beyond 2100 m asl. On shady south-
facing slopes C
4
grasses extend only up to ca. 2100 m asl. On
east- and west-facing slopes the transition lies between this
range (Parker et al., 2011). Extrapolating from data for
Lesotho published by Smith et al. (2002), the modern mean
maximum summer temperature at Sehonghong is ca. 26 ˚C,
where they assume a conservative lapse rate of 0.6 ˚C per
100 m [the dry adiabatic lapse rate is commonly approxi-
mated to 1 ˚C per 100 m (e.g. Tyson and Preston-Whyte,
2000)], and the modern average carbon isotope composition
of vegetation around the site is about 14‰. Thus, at 1870 m
asl Sehonghong lies well within the temperature envelope in
which C
4
grass taxa dominate today.
A complicating factor for reconstructing C
3
/C
4
grass distri-
butions beyond the Pleistocene/Holocene boundary is the
effect of lowered atmospheric CO
2
concentrations (pCO
2
)
during glacials. According to the model of Ehleringer et al.
(1997), lower pCO
2
conditions tend to favour C
4
plants
despite the lower temperatures. Thus, in glacial periods when
pCO
2
was demonstrably lower, altitudinal shifts of C
4
grasses
due to temperature shift would be partly suppressed. In effect,
this means that estimates of temperature depressions would
be conservative.
The average carbon isotope composition of the overlying
vegetation is recorded in the d
13
C values of proxies such as
soil organic matter (SOM) and herbivore tissues including
enamel. The isotopic composition of SOM is largely stable
after burial, and in sequence preserves a record of the
overlying biomass (Ehleringer et al., 2000). Many studies
have applied the stable isotope composition of SOM in
stratigraphic sequences to reconstruct changing C
3
/C
4
ratios
of the overlying vegetation (e.g. Tieszen et al., 1979; Parker
et al., 2011; Stewart et al., in press).
One important caveat for the present study relates to
sediments within a rockshelter. In these cases SOM does not
derive from plants growing in soil above, but rather from the
organic matter in sediments entering the site by wind, or
animal or human agency, and the organic material in the
form of food, fuel and grass-bedding brought into the site by
human occupants. We assume that grass, used abundantly for
bedding and fuel and evident in well-preserved mats within
the Pleistocene deposits (Mitchell, 1996b), is a major compo-
nent of SOM and that the grasses selected were representative
of those growing immediately around the rockshelter. The
other large contributors to SOM are likely to have been C
3
woody plants, which would tend to mask the changes in the
C
3
/C
4
grass ecotone. By contrast, an SOM sequence provides
a palaeoenvironmental archive that can be radiometrically
anchored, to act as a long-term, semi-continuous framework
for the site.
Similar principles related to altitudinally controlled vegeta-
tion shifts apply to the isotopic analysis of herbivore tooth
enamel, except that the animals’ behaviour and preferences
introduce an element of selection. Thus, the teeth of grazers
reliably record the isotopic composition of the grassy vegeta-
tion. Enamel is an incremental tissue laid down over 1–3
years and so it reflects the dietary isotope composition from
the time of tooth formation only, and is thus of comparatively
short duration unlike many other types of palaeoenvironmen-
tal archive, which may aggregate the environmental record
over years or decades. d
13
C of bone and enamel apatite
among large herbivores is typically about 12–14‰higher
than the plants they are eating (Lee-Thorp et al., 1989;
Cerling and Harris, 1999). Thus, reliable modern C
4
feeders
have a predicted range of 0 to þ2‰while C
3
feeders are
predicted to have carbon isotopic values of 12 to 14‰.
As grazers migrate seasonally in search of pastures, their teeth
reflect d
13
C of grassy vegetation regionally, and not necessar-
ily that locally around the site.
In the tropics and sub-tropics, the oxygen isotope composi-
tion of herbivore tooth enamel provides information on
d
18
O values in environmental water and plant water, as these
are the main sources of d
18
O variability in mammalian blood
bicarbonate (and enamel) (Ambrose and Norr, 1993; Tieszen
and Fagre, 1993). Thus, they provide an indirect reflection of
precipitation d
18
O values, hence allowing information about
moisture sources, storm tracks, and amounts and the nature
of precipitation (intense rain, snowfall, etc.). Interpreting
these archives is rarely straightforward because of the number
of variables influencing d
18
O values in meteoric water and
the proxy archive. Of relevance here is the pronounced
18
O depletion of precipitation falling as snow, and the
negative correlation between d
18
O and precipitation amount.
Temperature correlates positively with d
18
O, reflecting both
altitudinal/temperature and seasonality effects, among others
(Gat, 1996). The long-term weighted mean of modern
d
18
O of rainfall in southern Africa is 2.8‰(IAEA/WMO,
2006), but a more detailed view of the precipitation around
Sehonghong is provided by modelled data obtained from the
Online Isotopes in Precipitation Calculator (OIPC, water-
isotopes.org, accessed January 2015). The OIPC result, which
models the long-term average precipitation d
18
O at the site
coordinates and altitude of Sehonghong based on global
precipitation isotope data, is considerably more negative at
5.6‰, although the accuracy of such a global model at a
point scale requires verification (Bowen and Revenaugh,
2003; Bowen, 2015).
Oxygen isotope ratios of mammal tissues are influenced by
climate, diet and physiology. Oxygen is incorporated into
animal tissues in equilibrium with body water, and is derived
mainly from drinking water and water in food (Sponheimer
and Lee-Thorp, 1999a). While water available for drinking is
isotopically similar to meteoric water, bar any large evapora-
tion effects, d
18
O values of food water reflect both meteoric
water d
18
O and fractionation processes by plants. Thus,
Figure 2. Schematic diagram showing the relationship between
vegetation, altitude and temperature. Note that the vegetation zones
are higher on the warmer, north-facing slope. Adapted from Roberts
et al. (2013).
Copyright #2015 John Wiley & Sons, Ltd. J. Quaternary Sci. , Vol. 9999(9999) 1–12 (2015)
JOURNAL OF QUATERNARY SCIENCE
species that are obligate drinkers provide better records of
meteoric water values than species that drink little and obtain
most water from food (Bocherens et al., 1996; Kohn, 1996).
Unpublished data referred to in Smith et al. (2002) show
that d
18
O values in ungulate herbivores in the Limpopo
Province, South Africa, to the north of Sehonghong, increase
from 1.1 1.9‰(n¼11) to 3.1 1.4‰(n¼33) as aridity
and evaporation rates increase along a precipitation transect.
Closer to Sehonghong, faunal samples from two levels
dating to within the last 1000 years at Rose Cottage Cave
(ca. 1600 m asl) in the Caledon River Valley have mean
d
18
O values of 3.0 1.6‰(n¼5) and 4.3 1.5‰(n¼11)
(Smith et al., 2002).
Here we present results of carbon isotope analyses of SOM
from layers spanning from late MIS 3 ca. 35 ka to the mid-
Holocene, as well as more targeted carbon and oxygen
isotopic analyses of archaeological tooth enamel of grazing
herbivores from layers dated to late MIS 3 (ca. 35–33 ka). The
SOM sequence provides a semi-continuous assessment of
vegetation changes around Sehonghong, while the enamel
results contribute information on regional vegetation and
hydrological conditions for a poorly characterized period of
MIS 3. Together, these results provide complementary insights
into the vegetation changes over the period of occupation.
This reconstruction is anchored by nine new accelerator mass
spectrometry (AMS) radiocarbon dates on ABOx-SC pre-
treated charcoals, which extend the chronology into levels
recently excavated by AMEMSA.
Materials and methods
Radiocarbon dates
Previously published radiocarbon dates from Sehonghong
span the LSA through the transitional LSA/MSA levels
(Table 1). The recent AMEMSA excavations in Mitchell’s
trench extend into the MSA levels below, previously only
poorly constrained by four OSL ages from Carter’s trench
(Jacobs et al., 2008). Nine charcoal samples were selected
from the recent excavations (see Table 2 for sample context),
typically intact pieces ca. 1 cm in diameter, with only one
date derived from multiple fragments of charcoal (OxA-
27695). Samples were prepared in the University of Oxford
Research Laboratory for Archaeology and the History of Art
(RLAHA) Radiocarbon Unit according to the ABOx-SC (acid–
base-wet oxidation-stepped combustion) protocol, as the age
was expected to be >20 ka (Bird et al., 1999; Brock et al.,
2010). This technique is considerably more rigorous than
other pretreatment methods for charcoals and is more effec-
tive at removing contaminating carbon, particularly humic
acids, and generally produces older dates than less rigorous
pretreatment methods (Bird et al., 1999).
Approximately 180 mg of charcoal per sample was reacted
with 6 M HCl to demineralize the sample. Samples were
rinsed and reacted twice with 30 mL of 1 M NaOH, rinsing
between treatments, and subsequently treated with 30 mL of
an oxidizing solution of 0.1 m potassium dichromate in 2 m
sulphuric acid in a heating block at 60 ˚C for 18 h. The
remaining oxidation-resistant elemental carbon (OREC) was
repeatedly rinsed until the water ran clear, frozen and freeze-
dried overnight. Approximately 20 mg of OREC was heated in
the presence of oxygen to 630 ˚C for 2 h. Then, 3–3.5 mg of
OREC was measured for stable isotope composition and
carbon and nitrogen content in a CHN elemental analyser
(Carlo-Erba NA, 2000) coupled to a gas source isotope ratio
mass spectrometer (Sercon 20/20), with blank tins in between
each sample (Brock et al., 2010). Blank tins contain <2mgof
carbon. The CO
2
gas produced by samples was collected for
graphitization at 560 ˚C according to Dee and Bronk Ramsey
(2000). The resultant graphite powder was pressed into targets
and dated on the Oxford Radiocarbon Accelerator Unit’s
HVEE AMS system (described in Bronk Ramsey et al., 2004).
Stable light isotopes in SOM
Twenty-nine bulk soil samples were selected for isotopic
analysis, spanning 1.2 m of deposit, and taken with reference
to the natural stratigraphy (M. Morley, pers. comm.). Samples
were prepared according to a standard methodology (Cerling,
1991). The sediment was sieved through a 1 mm mesh to
remove charcoal fragments and other plant remains. Samples
with large amounts of charcoal were avoided to limit the
influence of woody C
3
plants on SOM d
13
C. Approximately
1 g was reacted for several days with 2 M HCl to remove
carbonates. Once the reaction had ceased, the samples were
rinsed repeatedly until the water ran clear and all fine
charcoal fragments had been poured off. Samples were
freeze-dried for 48 h. The dry sediments were weighed into
tin capsules, loaded into an automated Sercon GSL elemental
analyser and combusted to produce CO
2
. The evolved gases
(N
2
was not measured) were introduced to a Sercon Geo 20/
20 isotope ratio mass spectrometer, in continuous flow mode
using helium as a carrier gas.
Results were normalized according to the multi-point
method using CH-6 (sucrose; d
13
C¼10.449 0.033‰), an
international standard, and two in-house standards, alanine
(d
13
C¼26.91‰) and BROC-2 (powdered broccoli: d
13
C¼
27.48‰), which have been calibrated against international
standards (NBS-18, NBS-19 and NBS-22) (Paul et al., 2007).
These standards were chosen to provide a range of values
that were expected to bracket the range of sample values
(Paul et al., 2007). Results are reported in the delta-notation
relative to V-PDB, according to the equation d
a
X
(‰)¼{(R
sample
)/(R
standard
)–1}1000, where R
sample
is the
Table 2. ABOx-SC radiocarbon samples (charcoals) context description and sample d
13
C.
OxA Square Context Context description d
13
C(‰VPDB)
OxA-27689 J12 162 Rock fall in dark brown sandy silt matrix 25.1
OxA-27690 J12 163 Rock fall in pale brown sandy silt matrix 22.9
OxA-27691 J12 167 Thick burning horizon 22.0
OxA-27692 J12 169 Thick burning horizon 21.9
OxA-27693 J12 1030 Charcoal-rich layer 22.6
OxA-27694 J12 1031 Mottled black/strong brown sandy silt 22.8
OxA-27695 J12NE 1036 Rock fall in grey brown sandy silt matrix 21.9
OxA-27696 J12SE 1037 Hearth 24.5
OxA-27697 J12NE 1111A Thin charcoal artefact-rich layer 23.2
Copyright #2015 John Wiley & Sons, Ltd. J. Quaternary Sci. , Vol. 9999(9999) 1–12 (2015)
SEHONGHONG ROCKSHELTER PALAEOENVIRONMENTS
ratio of
a
X(the more abundant isotope) to
b
Xof the sample
material and R
standard
is that of a standard reference material.
d
13
C errors were <0.1‰for standards. However, intra-
sample uncertainty is probably higher for the SOM samples,
which are composed of a heterogeneous mix of organic
matter. Thus, several samples were run in duplicate or
triplicate, and averaged values are presented here. Standard
deviations between samples from the same layer are typically
good, on the order of 0.1–0.2‰, but in some instances can
be as high as 0.4‰. Thus, the results must be interpreted
cautiously.
Stable light isotopes in faunal enamel
The faunal teeth from the MSA levels at Sehonghong are
typically heavily fragmented, limiting possibilities for identifi-
cation. Twenty-nine teeth from newly dated strata were
identifiable to species and suitable for analysis. The species
consisted of two large reliable grazing species, hartebeest
(Alcelaphus buselaphus) and zebra (Equus sp.), and one
mixed feeder, eland (Taurotragus oryx). Because the enamel
was fragile, fragments were removed manually lengthwise
down the tooth using tweezers, averaging any isotopic
variability attributable to seasonal changes in d
13
C and
d
18
O during growth. Fragments were ground reacted in a 1: 4
(v/v) mixture of 7% sodium hypochlorite (NaOCl) and
distilled water for 30 min, washed to neutrality and then
reacted with 0.1 M acetic acid for 10 min (Sponheimer and
Lee-Thorp, 1999a,b). The samples were centrifuged and
washed several times between and after each reaction, before
they were freeze-dried overnight.
Enamel samples were analysed in one of two systems – a
VG Isocarb common acid bath preparation system attached
to a VG Isogas Prism II mass spectrometer in the University of
Oxford Earth Science Stable Isotope Laboratory, or an
automated Thermo GasBench II device, coupled to a Thermo
Delta V Advantage mass spectrometer at the Division of
Archaeological, Geographical and Environmental Sciences of
the University of Bradford. In each case CO
2
was produced
by 100% phosphoric acid hydrolysis. The samples were
calibrated against internal standards calibrated in turn to
international standards (NBS19: d
13
C¼1.95‰,d
18
O¼
2.20‰; NOCZ: d
13
C¼2.307‰,d
18
O¼1.906‰; and
OES: d
13
C¼10.52‰,d
18
O¼5.20‰in Oxford and
Merck-CaCO3: d
13
C¼35.25‰,d
18
O¼13.15‰; and CO-
1: d
13
C¼2.492 0.030‰,d
18
O¼2.4 0.1‰in Bradford).
Reproducibility for the Prism was 0.3‰for both d
13
C and
d
18
O, and for the Gasbench system was 0.1‰for d
13
C and
0.3‰for d
18
O. This level of reproducibility does not
significantly affect palaeoclimatic interpretations of the order
undertaken here.
Results and discussion
ABOx-SC radiocarbon dates
Table 1 presents the radiocarbon measurement and uncali-
brated date for each AMS sample, in sequence with the
previously published dates from Sehonghong. The dates
undertaken as part of this study (OxA-27689–97) range from
25 330 130 to 31 030 250
14
C a BP and fit well with the
radiocarbon chronology for overlying strata (Carter and Vogel,
1974; Carter, 1976; Carter et al., 1988; Mitchell and Vogel,
1994). All The charcoal samples selected survived ABOx-SC
treatment, attesting to the excellent preservation conditions at
Sehonghong: a project to date similarly aged charcoal samples
at the nearby Melikane archaeological site had much lower
survival rates (Wheeler, 2010; Stewart et al., 2012).
Table 1 also shows the calibrated age range for all
previously published and new radiocarbon dates, calculated
with OxCal v. 4.2 using the SHCal13 curve (for the Southern
Hemisphere) (Hogg et al., 2013). Calibration shows that the
new AMS dates undertaken in this study span between
35 288–34 585 and 29 876–29 036 cal a BP. Two dates,
OxA-27694 and OxA-27697, are out of stratigraphic order,
although the calibrated age ranges overlap with those above
or below.
Figure 3 shows a Sequence model based on Bayesian
statistics of all radiocarbon dates from Sehonghong (Bronk
Ramsey, 2009a), which defines the order in which the
samples were deposited. The model includes both Phases for
different cultural units (i.e. Wilton, Robberg, etc.) and a
nested P_Sequence model for the MSA AMS dates, for which
depth information is derived from comparison with the
unpublished section drawing for the north wall of square J12
(P. Mitchell, pers. comm.). Six outliers were excluded from
the larger model according to the agreement index method
(Bronk Ramsey, 2009b). Compared with the unmodelled
AMS dates in Table 1, the nested P_Sequence-transformed
AMS dates have a narrower range of probable values as they
are constrained by those above and below. Within the
P_sequence for the new AMS dates, the most recent AMS
date, OxA-27689 (29 876–29 036 cal a BP), is noticeably
offset from those below, and the change in depth suggests a
hiatus, or a slowed rate of deposition below this determina-
tion. Conversely, a period of relatively rapid deposition
appears to have taken place between OxA-27695 (34 934–
34 226 cal a BP) and OxA-27694 (33 739–33 201 cal a BP),
which spans a considerable section of deposit (note that the
P_Sequence model has a noticeable effect on the probable
value of one date in particular, OxA-27694, which is shifted
towards the older values below).
These new dates extend the detailed radiocarbon chronol-
ogy for Sehonghong back to ca. 35 ka (the OSL dates,
taken from Carter’s trench excavated to bedrock, indicate
the deposits extend to the beginning of MIS 3, but provide
low chronological coverage for these earlier levels). Taken
together (and calibrated), the new and old radiocarbon dates
suggest the site was visited in recurrent occupational pulses
through late MIS 3 (ca. 35–32 ka and ca. 30–29 ka), early MIS
2 (ca. 25–21 ka and ca. 19–18.5 ka), mid/late MIS 2 (ca.
16–12.5 ka), the Pleistocene/Holocene transition (ca. 11.5–
10 ka), and the early (ca. 8–7.5 ka), mid (ca. 7–6.5 ka) and
late (ca. 1.7–1 ka) Holocene (Fig. 3). The temporal gaps
between occupational pulses may represent periods that saw
little or no human habitation or sediment deposition, as
archaeologically sterile deposits were not encountered. More
extensive dating efforts may change this picture, however, just
as new dating samples from as yet unexcavated sediments in
Mitchell’s trench should extend the radiocarbon chronology
further back in time.
SOM data
d
13
C results for the SOM samples aged ca. 7–35 ka are
presented in Table 3, while Fig. 4 shows d
13
C changes with
depth. Approximate ages in Table 3 are assigned to samples
based on depth information and the unpublished section
drawing for the north wall of square J12 (P. Mitchell, pers.
comm.). Values range from 19.3 to 24.7‰. These data
reflect a large C
3
component to the vegetation or organic
matter brought into the site until the end of the Pleistocene.
Before the start of the Holocene, values for SOM (mean ¼
23.7 0.5‰,n¼20) lie consistently close to the predicted
value of ca. 25‰for 100% C
3
vegetation before the
Copyright #2015 John Wiley & Sons, Ltd. J. Quaternary Sci. , Vol. 9999(9999) 1–12 (2015)
JOURNAL OF QUATERNARY SCIENCE
influence of fossil fuels on atmospheric CO
2
. This pattern
suggests temperatures were too low during this late phase of
the Last Glacial for C
4
taxa to extend to the altitude of the
rockshelter at ca. 1870 m asl. The modern maximum altitudi-
nal range for C
4
grassland on north- and south-facing slopes
is ca. 2700 and 2100 m asl, respectively, and between that
range on east- and west-facing slopes (Parker et al., 2011).
Sehonghong faces north-west, and we assume people col-
lected grasses from several hundred metres around the site,
averaging micro-scale variability. Applying a 0.6 ˚C per
100 m temperature–altitude relationship for the distribution of
C
4
grasses, the predominantly C
3
signal indicates that temper-
atures were at least 5 ˚C lower during the Pleistocene. This
conclusion is not affected by uncertainties over the effect of
lowered pCO
2
during the Last Glacial, which would have
favoured C
4
vegetation expansion upslope, and is thus a
conservative estimate of the temperature depression, which
may have been several degrees greater (Ehleringer et al.,
1997). While a minimum figure, this estimate is consistent
with estimates for a 3–5 ˚C decrease in temperature at the
LGM based on stable isotope data from slightly lower
elevation sites in western Lesotho (Smith et al., 2002). The
charcoal assemblage at Melikane just before the LGM shows
an increase in frost-tolerant taxa and reduced tree cover,
possibly restricted to sheltered river valleys (Stewart et al., in
press). Evidence for limited glaciation in the Maloti–Drakens-
berg around the LGM similarly points to considerably
reduced temperatures year-round (Mills et al., 2009).
The upper section of the sequence shows considerably
more variability. SOM d
13
C shifts to more positive values,
indicating a warming trend near the Pleistocene/Holocene
boundary, identified at the SA/RBL transition. Notably,
Figure 3. Bayesian model of calibrated radiocarbon dates from Sehonghong, consisting of multiple Phases and a nested P_Sequence model for
the AMS dates. The white distributions are the original age estimates, while the black distributions are the modelled results. Calibrated and plotted
in OxCal v. 4.2, using SHCal13.
Copyright #2015 John Wiley & Sons, Ltd. J. Quaternary Sci. , Vol. 9999(9999) 1–12 (2015)
SEHONGHONG ROCKSHELTER PALAEOENVIRONMENTS
there is a sharp increase in d
13
C(19.7 0.4‰,n¼5)
suggesting a greater contribution of C
4
taxa at ca. 14 ka.
A return to C
3
-dominated conditions is indicated by a
series of more negative d
13
C values at ca. 11 ka (mean ¼
21.9 0.2‰,n¼5), and followed by consistently more
positive values in the early to middle Holocene layers. This
section reflects the establishment in the area of larger
proportions of C
4
grasses under relatively warm conditions
(mean ¼19.7 0.4‰,n¼3). The difference between the
average values for the early to mid-Holocene levels and the
Pleistocene levels is 4.0‰. These results are consistent with
the d
13
C results of SOM from two lower-lying (ca. 1640 m
asl) sites in the region, Ha Makotoko and Ntloana Tsoana
(Roberts et al., 2013), which record very rapid swings in
grassland composition at the Pleistocene/Holocene transition,
probably reflecting the global temperature fluctuations at this
time. Roberts et al. (2013) report a sharp warming trend at ca.
11 ka, with an average increase of 3.7‰, followed by a
return to slightly more negative values. Similarly, Smith et al.
(2002) reported a cool reversal of the general warming trend
in the eastern Free State at ca. 10.2–9.5 ka (uncalibrated),
which they suggest may represent an extension of the
Younger Dryas cooling event observed in southern Africa ca.
11 ka. The differences in timing observed between these
studies may be explained by differences in dating methods
and their precision, and by the different altitudes of the sites,
as higher elevations show more sensitivity to temperature
changes (Scott, 1989; Smith et al., 2002). Notwithstanding,
the Sehonghong SOM record is consistent with emerging
reconstructions of relatively large (ca. 3–5 ˚C) temperature
swings in southern Africa at the Pleistocene/Holocene bound-
ary between ca. 14 and 11 ka.
Herbivore enamel data
The teeth sampled come from layers dated from 35 105–34 225
to 33 420–31 917 cal a BP, based on the radiocarbon ages
presented above, and so span only a relatively short period of
(at most) 3000 years. Nonetheless, the d
13
C values range from
11.1 to 5.8‰(mean ¼8.1 1.4‰;median¼8.3‰)
across this period, supporting reconstructions of climatic vari-
ability during MIS 3. Table 4 presents the stable isotope results
for the teeth with context and associated radiocarbon dates, and
Fig. 5 shows d
13
Candd
18
O results in stratigraphic sequence.
Where there are too few samples per layer, levels have been
aggregated. Also shown are boxplots of all values per layer. No
significant differences are observed between levels for average
d
13
Candd
18
O values of grazing species.
Ideally, any environmental reconstruction should incorpo-
rate results from several individuals of several species with
Table 3. Stable carbon isotope data from soil organic matter in the Sehonghong stratigraphic profile, showing depth, stratigraphic association,
associated radiocarbon date reference and calibrated age. Where duplicate or triplicate measurements were taken the average is shown, with the
standard deviations and % carbon (calculated from the IRMS runs) shown alongside.
Approx. date (cal a BP)†
Sample Maximum depth (cm) Stratigraphic contextDate ID From To d
13
C(‰VPDB) SD (‰VPDB) % C
SOM3 17 GWA Pta-6154 6931 6548 20.3 0.23 1.4
SOM4 19 19.3 5.3
SOM6 28 SA Pta-6368 10 550 10 225 19.5 0.03 0.8
SOM8 32 22.2 0.7
SOM9 34 21.8 0.05 10.6
SOM10 36 21.7 5.3
SOM11 38 Pta-6057 11 595 10 588 22.0 0.10 11.8
SOM12 40 SA-RBL 21.8 0.17 25.0
SOM12a 42 RBL Pta-6062 14 467 13 731 19.8 0.43 8.9
SOM13 44 22.6 0.20 5.3
SOM15 53 23.7 4.4
SOM16 57 CLBRF Pta-6058 15 040 14137 23.8 0.09 18.5
SOM17 59 BAS Pta-6060 19 298 18 600 23.2 0.15 2.7
SOM18 62 23.6 0.43 2.5
SOM20 68 Pta-6077 24 495 23 952 22.9 1.5
SOM21 70 162 Pta-6059 25 272 24 061 23.6 2.9
SOM22 72 23.8 0.6
SOM23 79 24.2 0.09 42.5
SOM24a 81 163 Pta-6271 29 890 28 466 23.7 9.7
SOM24b 83 Pta-6268 30 983 29 243 24.5 3.5
SOM24c 84 OxA-27690 33 148 31 917 24.7 0.05 1.4
SOM26 95 166 OxA-27691 33 420 32 765 23.3 0.05 11.8
SOM27 99 167 23.5 0.37 22.3
SOM29 109 169 OxA-27692 33 531 32 966 24.3 0.12 14.7
SOM32 116 1031 OxA-27693 33 630 33 095 23.5 3.5
SOM34 119 1032 upper 24.0 4.8
SOM35 123 1032 lower 23.7 4.9
SOM38 129 1035 OxA-27695 34 934 34 226 24.0 0.10 4.4
SOM40 137 1036 23.0 3.6
The associated stratigraphic context is based on the unpublished profile of the J12 north wall (P. Mitchell, pers. comm. and unpublished fieldnotes, M.
Morley, pers. comm.).
†The approximate dates are assigned on the basis of proximity to a radiocarbon-dated sample.
Pretoria radiocarbon dates on charcoal are from Mitchell (1996b), OxA dates from this study. The SHCal13 (Hogg et al., 2013) calibration curve
was used; calibrated ages are presented at 95% confidence interval.
Copyright #2015 John Wiley & Sons, Ltd. J. Quaternary Sci. , Vol. 9999(9999) 1–12 (2015)
JOURNAL OF QUATERNARY SCIENCE
varying diets and physiologies. However, this is difficult to
achieve where the number of well-preserved and identifiable
elements is very low, as in many southern African MSA
archaeological contexts. Interpretation of palaeoclimatic and
environmental variables using faunal stable isotope values
therefore has to proceed with an awareness of how animal
behaviour, physiology and diet filter the environmental
signal. All three species are predictable indicators of past
grassland composition and hydrological conditions. Neither
zebra nor red hartebeest are thought to incorporate significant
amounts of browse into their diets. Eland, however, are
known to be mixed feeders at times, although they are
primarily grazers. Both zebra and hartebeest must drink
regularly (Kingdon, 1997), which means the
18
O values of
their teeth largely reflect the d
18
O value of available drinking
water and precipitation (Bocherens et al., 1996; Kohn, 1996).
Eland, however, derive much of their bodywater from their
diet, although they do drink when water is available (Pappas,
2002). Thus, the oxygen isotope composition of their enamel
might be expected to be more enriched in
18
O. However, no
statistically significant differences between the species are
observed.
The d
13
C results suggest that most of the animals sampled
here ate substantial proportions of C
4
grass (several individu-
als with d
13
C values of 6.5 to 7‰imply contributions of
up to 40% C
4
plants), consistent with historical observations
that they migrated seasonally between the uplands and
lowveld. However, the range of values is considerably more
negative than average values reported from Late Pleistocene
Table 4. Enamel d
13
C and d
18
O results presented in stratigraphic order with archaeological context. Most of the analyses were carried out in the
University of Oxford’s Earth Science Stable Isotope Laboratory; a subset (indicated ) were analysed in the University of Bradford’s stable isotope
facility (as indicated in Methods).
Approx. date (cal a BP)
Species Square Context d
13
C(‰VPDB) d
18
O(‰VPDB) From To
EquidJ13 163 6.9 1.9 33148 31 917
Equid K12 166 8.5 2.9
Equid K12 167 8.1 2.6 33 420 32 765
Alcelaphine K12 167 7.1 1.9
Equid K12 168–169 11.1 6.8
EquidK12 168–169 10.2 2.8
EquidJ12 169 9.0 0.7 33 531 32 966
EquidJ12 169 11.1 3.1
Equid J12 169 6.8 0.3
Equid J12 169 8.3 2.2
Equid J12 169 6.5 0.1
Alcelaphine I13 166–170 6.1 3.2
Equid J12 170A 8.8 0.7
Eland J12 1030 8.6 3.9 33 630 33 095
Eland J12 1031 8.9 2.9 33 739 33 201
Alcelaphine J12 1031 8.7 0.1
Equid J12 1031 9.0 0.1
Equid J12 1033 8.2 0.4
Equid J13 1023 6.5 0.1
Alcelaphine J13 1024 8.5 0.8
Equid J13 1024 5.9 2.3
Alcelaphine I12 1005 8.5 1.7
Alcelaphine J12 1035 8.7 1.8 34 934 34 226
Equid K12NW 1045 7.9 0.4
Alcelaphine I12 1006 6.7 0.1 34 934 34 226
Alcelaphine I12 1006 7.6 0.2
Equid J13 1026 5.8 1.3
Equid I12NW 1009 6.9 0.4 35 105 34 485
Equid I12 1016 8.7 0.9
Calibrated dates are from this study, with 95% confidence interval.
Figure 4. Soil organic matter d
13
C (PDB ‰) with depth below the
surface (cm). The size of the markers reflects the range of uncertainty
of 0.2‰. Approximate ages are as determined from calibrated
radiocarbon dates.
Copyright #2015 John Wiley & Sons, Ltd. J. Quaternary Sci. , Vol. 9999(9999) 1–12 (2015)
SEHONGHONG ROCKSHELTER PALAEOENVIRONMENTS
enamel samples dated to ca. 16 ka from Rose Cottage
Cave and Tloutle of 2.5‰(n¼7) and 4.5‰(n¼2),
respectively (Smith et al., 2002). Most interesting are the three
samples with the most negative d
13
C values, which all occur
in layer 169, dated to 33 531–32 966 cal a BP. These
samples, with d
13
C between 11.1 and 10.2‰, lie close to
the value expected for a 100% C
3
diet (12 to 14‰) and
indicate that vegetation belts were substantially lowered at
this time, even compared with the cool conditions at the end
of the Pleistocene. This evidence for a rapid climate excur-
sion is consistent with a particularly negative d
18
O value
(replicate analyses of 6.5 and 7.0‰) for a single individ-
ual from layer 169 (see below).
The d
18
O results range from 6.8 to 1.9‰, with an
average value of 1.3 1.8‰(median ¼0.8‰), but there
is no strong patterning. In general, these values are consider-
ably more negative than Late Holocene samples from Rose
Cottage Cave [3.0 1.6‰(n¼5) and 4.3 1.5‰(n¼11)]
(Smith et al., 2002). They are also more negative than Late
Pleistocene samples from Rose Cottage Cave (0.6‰;n¼7)
and Tloutle (1.5‰;n¼2) dating to ca. 16 ka (Smith et al.,
2002). Assuming that the d
18
O values reflect meteoric water
values averaged for the entire region, the mean for the
Sehonghong fauna at ca. 35 ka compares well with the
average modern value of 1.14‰for faunal enamel from a
relatively mesic southern Africa environment (unpublished
data, referred to in Smith et al., 2002). Alternatively, the low
values may reflect greater amounts of precipitation falling in
the highlands, possibly as snow. In particular, the low mean
d
18
O value of 6.8‰from an equid tooth in layer 169 hints
that this animal may have received some of its drinking water
from meltwater, which is markedly depleted in
18
O. This is
supported by limited evidence for viable, if marginal glacia-
tion in the highest reaches of the Maloti–Drakensberg
Mountains during MIS 3 and at the LGM (Mills et al., 2009,
p. 647). Even marginal glaciation during MIS 3 would have
necessitated higher amounts of winter precipitation than have
been previously estimated (Mills et al., 2009).
Conclusions
ABOx-SC pretreated radiocarbon dates from ca. 35–30 ka
confirm human occupation of Sehonghong during a climati-
cally dynamic period. Although MIS 3 has long been thought to
have been a period of uniformly dry conditions across southern
Africa, with widespread depopulation of the interior (Deacon
and Thackeray, 1984), the palaeoenvironmental evidence has
been ambiguous, suggesting highly variable conditions across
MIS 3 (Scott, 1982; Partridge et al., 1997; Gasse et al., 2008).
Many more sites dating to MIS 3 have been discovered in the
interior region in recent years (Mitchell, 2008), suggesting that
populations adapted to the challenging climatic conditions,
possibly by employing flexible strategies, including seasonal
migration between uplands and lowlands (Carter, 1976), and
pulsed occupations in the uplands that coincide with periods of
regional heightened aridity and/or instability (Stewart et al.,
2012, in press). While numbers are limited, the d
18
Oresults
reported here for faunal enamel are consistent with a model in
which the highlands remained moist. The low d
18
O values,
particularly around 33 ka, are consistent with reconstructions
of greater amounts of precipitation, potentially falling as
18
O-depleted snow during winter. Thus, while the area was
undoubtedly much colder, the availability of abundant water
may well explain the continued attraction of Sehonghong as a
living site during this climatically challenging period.
Carbon isotope analyses of faunal enamel samples from a
brief time slice of late MIS 3 indicate a substantial lowering
of C
3
/C
4
vegetation belts. SOM d
13
C results extending to ca.
35 ka similarly indicate a depression of vegetation belts
during the Last Glacial, with a temperature difference from
today of at least 5 ˚C before the onset of warming at the end
of the Pleistocene. The SOM data support climatic recon-
structions that indicate that the transition from the Pleistocene
to the Holocene was a time of considerable climate variabil-
ity (Smith et al., 2002; Roberts et al., 2013). Rapid changes in
grassland composition recorded in the Sehonghong sequence
may reflect global temperature events, such as the Younger
Dryas stadial (ca. 12 800–11 500 BP).
Given reconstructions of climatic variability across south-
ern Africa during MIS 3, the reliable availability of water, and
related plant and faunal resources including freshwater fish,
may account for the presence of humans at the site at
this time of dramatically lowered temperatures (Mitchell,
1990; Stewart et al., 2012, in press). The nearly continuous
occupation of Sehonghong through late MIS 3 supports
conclusions about the resilience of Late Pleistocene hunter-
gatherers in the face of considerable environmental stress.
The results of this research will be complemented by
on-going faunal, sedimentological and phytolith analyses.
The secure chronology of the site, now extending back to ca.
35 ka, and the extraordinary preservation of organic material
is promising for the exploration of hunter-gatherer lifestyles in
this highland environment during phases of major environ-
mental and adaptive change.
Figure 5. d
13
C and d
18
O of enamel samples, plotted by aggregated archaeological context, with calibrated radiocarbon ages (Table 2). No
significant differences are observed between the three species, or between levels. Also shown are boxplots of the values for the two grazing
species per level (zebra and hartebeest only), with the mean indicated by the solid horizontal lines.
Copyright #2015 John Wiley & Sons, Ltd. J. Quaternary Sci. , Vol. 9999(9999) 1–12 (2015)
JOURNAL OF QUATERNARY SCIENCE
Acknowledgments. Funding was provided to E.L. by the University of
Oxford Clarendon Fund and Merton College. AMEMSA is supported
by grants from the McDonald Institute for Archaeological Research,
the University of Cambridge, the British Academy, the Wenner-Gren
Foundation, the Prehistoric Society, the Institute for Field Research,
and the Social Sciences and Humanities Research Council of Canada.
Permits to excavate at Sehonghong Rockshelter were generously
granted (to B.A.S.) by the Protection and Preservation Committee (PPC)
of the Lesotho Department of Culture. We thank bo-Mme Moliehi
‘Maneo Ntene, Puseletso Moremi, Matsosane Molibeli, Tsepang Shano
and the late Ntsema Khitsane for their support. Peter Ditchfield, Andy
Gledhill, Mike Morley and Katerina Douka provided valuable
assistance with laboratory analyses, and Peter Mitchell provided
insights into the depositional sequence and archaeology of the site.
Finally, we thank two anonymous reviewers whose suggestions helped
improve an earlier draft of this paper.
Abbreviations. AMEMSA, ‘Adaptations to Marginal Environments
during the Middle Stone Age’; AMS, accelerator mass spectrometry;
LGM, Last Glacial Maximum; LSA, Later Stone Age; MIS, Marine
Isotope Stage; MSA, Middle Stone Age; OREC, oxidation-resistant
elemental carbon; OSL, optically stimulated luminescence; SOM, soil
organic matter
References
Ambrose SH, Norr L. 1993. Experimental evidence for the relation-
ship of the carbon isotope ratios of whole diet and dietary protein
to those of bone collagen and carbonate. In Prehistoric Human
Bone: Archaeology at the Molecular Level, Lambert J, Grupe G
(eds). Springer: Berlin; 1–37. Heidelberg.
Bird M, Ayliffe LK, Fifield L et al. 1999. Radiocarbon dating of ‘old’
charcoal using a wet oxidation, stepped-combustion procedure.
Radiocarbon 41: 127–140.
Bocherens H, Koch PL, Mariotti A et al. 1996. Isotopic biogeochem-
istry (13C, 18O) of mammal enamel from African Pleistocene
hominid sites: implications for the preservation of paleoclimatic
signals. Palaios 11: 306–318 [DOI: 10.2307/3515241].
Bowen GJ. 2015. The online isotopes In precipitation calculator,
version 2.2. www.waterisotopes.org.
Bowen GJ, Revenaugh J. 2003. Interpolating the isotopic composition
of modern meteoric precipitation. Water Resources Research 39:
1299 [DOI: 10.1029/2003WR002086].
Brock F, Higham TFG, Ditchfield P et al. 2010. Current pretreatment
methods for AMS radiocarbon dating at the oxford radiocarbon
accelerator unit (ORAU). Radiocarbon 52: 103–112.
Bronk Ramsey C. 2009a. Bayesian analysis of radiocarbon dates.
Radiocarbon 51: 337–360.
Bronk Ramsey C. 2009b. Dealing with outliers and offsets in
radiocarbon dating. Radiocarbon 51: 1023–1045.
Bronk Ramsey C, Higham TFG, Leach P. 2004. Towards high-
precision AMS: progress and limitations. Radiocarbon 46: 17–24.
Carter P. 1976. The effects of climatic change on settlement in
Eastern Lesotho during the Middle and Later Stone Age. Archaeol-
ogy World 8: 197–206.
Carter PL, Mitchell PJ, Vinnicombe P. 1988. Sehonghong: the Middle
and Later Stone Age industrial sequence from a Lesotho rock-shelter.
Oxford: British Archaeological Reports International Series 406.
Carter PL, Vogel JC. 1974. The dating of industrial assemblages from
stratified sites in eastern Lesotho. Man 9: 557–570 [DOI: 10.2307/
2801124].
Cerling TE. 1991. Carbon dioxide in the atmosphere: evidence from
Cenozoic and Mesozoic paleosols. American Journal of Science
291: 377–400 [DOI: 10.2475/ajs.291.4.377].
Cerling TE, Harris JM. 1999. Carbon isotope fractionation between
diet and bioapatite in ungulate mammals and implications for
ecological and paleoecological studies. Oecologia 120: 347–363
[DOI: 10.1007/s004420050868].
Cowling RM. 1983. The occurrence of C
3
and C
4
grasses in fynbos
and allied shrublands in the south Eastern Cape, South Africa.
Oecologia 58: 121–127 [DOI: 10.1007/BF00384551].
Deacon HJ, Thackeray J. 1984. Late Pleistocene environmental
changes and implications for the archaeological record in southern
Africa. In Late Cainozoic Palaeoclimates of the Southern Hemi-
sphere, Vogel JC (ed.). Balkema: Rotterdam; 375–390.
Dee M, Bronk Ramsey C. 2000. Refinement of graphite target
production at ORAU. Nuclear Instruments and Methods in Physics
Research Section B: Beam Interactions with Materials and Atoms
172: 449–453 [DOI: 10.1016/S0168-583X(00)00337-2].
Ehleringer JR, Buchmann N, Flanagan LB. 2000. Carbon isotope
ratios in belowground carbon cycle processes. Ecological Applica-
tions 10: 412–422 [DOI: 10.1890/1051-0761(2000)010[0412:CIR-
IBC]2.0.CO;2].
Ehleringer JR, Cerling TE, Helliker BR. 1997: C4. Photosynthesis,
atmospheric CO
2
, and climate. Oecologia 112: 285–299.
Gasse F, Chali
e F, Vincens A et al. 2008. Climatic patterns in equatorial
and southern Africa from 30,000 to 10,000 years ago reconstructed
from terrestrial and near-shore proxy data. Quaternary Science
Reviews 27: 2316–2340 [DOI: 10.1016/j.quascirev.2008.08.027].
Gat JR. 1996. Oxygen and hydrogen isotopes in the hydrologic cycle.
Annual Review of Earth and Planetary Sciences 24: 225–262 [DOI:
10.1146/annurev.earth.24.1.225].
Hogg A, Hua Q, Blackwell PG et al. 2013. SHCal13 Southern
hemisphere calibration, 0–50,000 years cal BP. Radiocarbon 55:
1889–1903 [DOI: 10.2458/azu_js_rc.55.16783].
IAEA/WMO, 2006. Pretoria station: Global Network of Isotopes in
Precipitation. International Atomic Energy Agency/Worl Meteoro-
logical Association: Geneva.
Jacobs Z, Roberts RG, Galbraith RF et al. 2008. Ages for the Middle
Stone Age of southern Africa: implications for human behavior and
dispersal. Science 322: 733–735 [DOI: 10.1126/science.1162219].
Kingdon J. 1997. The Kingdon Field Guide to African Mammals.
Academic Press: London.
Kohn MJ. 1996. Predicting animal d
18
O: accounting for diet and
physiological adaptation. Geochimica et Cosmochimica Acta 60:
4811–4829 [DOI: 10.1016/S0016-7037(96)00240-2].
Lee-Thorp JA, Sealy JC, Van Der Merwe NJ. 1989. Stable carbon
isotope ratio differences between bone collagen and bone apatite,
and their relationship to diet. Journal of Archaeological Science 16:
585–599 [DOI: 10.1016/0305-4403(89)90024-1].
Lee-Thorp JA, Talma SA. 2000. Stable light isotopes and environ-
ments in the southern African Quaternary and Late Pliocene. In
The Cenozoic of Southern Africa, Partridge TC, Maud R (eds).
Oxford University Press: Oxford; 236–251.
Lombard M, Wadley L, Deacon J et al. 2012. South African and
Lesotho Stone Age sequence updated (I). South African Archaeo-
logical Bulletin 67: 120–144.
Mills SC, Grab SW, Carr SJ. 2009. Recognition and palaeoclimatic
implications of late Quaternary niche glaciation in eastern Lesotho.
Journal of Quaternary Science 24: 647–663 [DOI: 10.1002/jqs.1247].
Mitchell P. 1990. A palaeoecological model for archaeological site
distribution in southern Africa during the Upper Pleniglacial and
Late Glacial. In The World at 18,000 BP: Low Latitudes, Gamble
C, Soffer O (eds). Unwin Hyman: London.
Mitchell P. 1995. Revisiting the Robberg: new results and a revision
of old ideas at Sehonghong rock-shelter, Lesotho. South African
Archaeological Bulletin 50: 28–38.
Mitchell P. 1996a. The late Quaternary landscape at Sehonghong in
the Lesotho highlands, southern Africa. Antiquity 70: 623–638.
Mitchell PJ. 1996b. The Late Quaternary of the Lesotho Highlands,
southern Africa: preliminary results and future potential of ongoing
research at Sehonghong shelter. Quaternary International 33:
35–43 [DOI: 10.1016/1040-6182(95)00097-6].
Mitchell PJ. 1996c. Sehonghong: the late Holocene assemblages with
pottery. South African Archaeological Bulletin 51: 17–25 [DOI:
10.2307/3888928].
Mitchell P. 2008. Developing the archaeology of marine isotope
stage 3. South African Archaeological Society Goodwin Series 10:
52–65.
Mitchell P, Vogel JC. 1994. New radiocarbon dates from Sehonghong
rock-shelter, Lesotho. South African Journal of Science 90: 284–288.
Pappas LA. 2002. Taurotragus oryx.Mammalian Species 689: 1–5
[DOI: 10.1644/1545-1410(2002)689<0001:TO>2.0.CO;2].
Parker AG, Lee-Thorp JA, Mitchell PJ. 2011. Late Holocene Neoglacial
conditions from the Lesotho highlands, southern Africa: phytolith
and stable carbon isotope evidence from the archaeological site of
Copyright #2015 John Wiley & Sons, Ltd. J. Quaternary Sci. , Vol. 9999(9999) 1–12 (2015)
SEHONGHONG ROCKSHELTER PALAEOENVIRONMENTS
Likoaeng. Proceedings of the Geologists’ Association 122: 201–211
[DOI: 10.1016/j.pgeola.2010.09.005].
Partridge TC, Demenocal PB, Lorentz SA et al. 1997. Orbital forcing
of climate over South Africa: a 200,000-year rainfall record from
the Pretoria Saltpan. Quaternary Science Reviews 16: 1125–1133
[DOI: 10.1016/S0277-3791(97)00005-X].
Paul D, Skrzypek G, F
orizs I. 2007. Normalization of measured stable
isotopic compositions to isotope reference scales a review. Rapid
Communications in Mass Spectrometry 21: 3006–3014 [DOI:
10.1002/rcm.3185].
Plug I. 1997. Late Pleistocene and Holocene hunter-gatherers in the
Eastern Highlands of South Africa and Lesotho: a faunal interpreta-
tion. Journal of Archaeological Science 24: 715–727 [DOI:
10.1006/jasc.1996.0153].
Plug I, Mitchell P. 2008. Sehonghong: hunter-gatherer utilization of
animal resources in the highlands of Lesotho. Annals of the
Transvaal Museum 45: 31–53.
Roberts P, Lee-Thorp JA, Mitchell PJ et al. 2013. Stable carbon
isotopic evidence for climate change across the late Pleistocene to
Early Holocene from Lesotho, southern Africa. Journal of Quater-
nary Science 28: 360–369 [DOI: 10.1002/jqs.2624].
Scott L. 1982. A Late Quaternary pollen record from the Transvaal
bushveld, South Africa. Quaternary Research 17: 339–370 [DOI:
10.1016/0033-5894(82)90028-X].
Scott L. 1989. Climatic conditions in southern Africa since the last
glacial maximum, inferred from pollen analysis. Palaeogeography
Palaeoclimatology Palaeoecology 70: 345–353 [DOI: 10.1016/
0031-0182(89)90112-0].
Smith JM, Lee-Thorp JA, Sealy JC. 2002. Stable carbon and oxygen
isotopic evidence for late Pleistocene to middle Holocene climatic
fluctuations in the interior of southern Africa. Journal of Quaternary
Science 17: 683–695 [DOI: 10.1002/jqs.687].
Sponheimer M, Lee-Thorp JA. 1999a. Oxygen isotopes in enamel
carbonate and their ecological significance. Journal of Archaeolog-
ical Science 26: 723–728 [DOI: 10.1006/jasc.1998.0388].
Sponheimer M, Lee-Thorp JA. 1999b. Alteration of enamel carbonate
environments during fossilization. Journal of Archaeological Sci-
ence 26: 143–150 [DOI: 10.1006/jasc.1998.0293].
Stewart BA, Dewar GI, Morley MW et al. 2012. Afromontane foragers
of the Late Pleistocene: Site formation, chronology and occupa-
tional pulsing at Melikane Rockshelter, Lesotho. Quaternary
International 270: 40–60 [DOI: 10.1016/j.quaint.2011.11.028].
Stewart BA, Parker AG, Dewar GI, Morley M, Allott L. In press.
Follow the Senqu: Maloti–Drakensberg paleoenvironments and
implications for early human dispersals into mountain systems. In
Africa from MIS 6-2: Population Dynamics and Paleoenvironments,
Jones S, Stewart BA (eds). Springer: Cham, Switzerland.
Teeri JA, Stowe LG. 1976. Climatic patterns and the distribution of
C4 grasses in North America. Oecologia 23: 1–12 [DOI: 10.1007/
BF00351210].
Tieszen LL, Fagre T. 1993. Effect of diet quality and composition on
the isotopic composition of respiratory CO
2
, bone collagen,
bioapatite, and soft tissues. In Prehistoric Human Bone: Archaeol-
ogy at the Molecular Level, Lambert J, Grupe G (eds). Springer:
Berlin; 121–155.
Tieszen LL, Senyimba MM, Imbamba SK et al. 1979. The distribution
of C3 and C4 grasses and carbon isotope discrimination along an
altitudinal and moisture gradient in Kenya. Oecologia 37: 337–350
[DOI: 10.1007/BF00347910].
Tyson PD, Preston-Whyte R. 2000. The weather and climate of
southern Africa. Oxford University Press: Cape Town.
Vogel JC, Fuls A, Ellis R. 1978. The geographical distribution of Kranz
grasses in South Africa. South African Journal of Science 74:
209–219.
Von Richter W. 1971. Past and present distribution of the black
wildebeest, Connochaetes gnou Zimmermann (Artiodactyla: Bovi-
dae) with special reference to the history of some herds in South
Africa. Annals of the Transvaal Museum 27 35: 47.
Wheeler M. 2010. Radiocarbon dating of Melikane Rockshelter.
MSc Thesis, University of Oxford.
Copyright #2015 John Wiley & Sons, Ltd. J. Quaternary Sci. , Vol. 9999(9999) 1–12 (2015)
JOURNAL OF QUATERNARY SCIENCE