7.3.1 Denuding landscapes

The age of a landscape that is undergoing erosion is notional: a finite erosion rate remains finite however short the measurement time, and an eroding landscape is therefore always infinitesimally young. Instead of reporting an age, it is more accurate to calculate the rate of denudation, and cosmogenic dating has been applied to quantify this rate. Kounov et al. (Reference Kounov, Niedermann, de Wit, Viola, Andreoli and Erzinger2007) report denudation rates for the interior of South Africa of 1.5–3.0 m/Ma while Decker et al. (Reference Decker, Niedermann and de Wit2011) also report rates of <4.0 m/Ma. Landscapes that denude at >5 m/Ma are typically subject to in situ weathering rather than fluvial erosion (Baldwin et al., Reference Baldwin, Whipple and Tucker2003). This is supported by Decker et al. (Reference Decker, Niedermann and de Wit2013) who compared denudation rates for Karoo dolerite surfaces of <4.0 m/Ma using a climate-dependent weathering model and concluded that high weathering rates limit landscape denudation. However, Decker et al. (Reference Decker, Niedermann and de Wit2011) reviewed the available data for erosion rates and concluded that weathering is presently happening at a rate that is orders of magnitude slower than the rate of soil loss. These results indicate that chemical weathering and physical erosion of the southern African landscape must be closely matched on a very slowly denuding land surface. This is illustrated in Kruger National Park where the denudation rate across a rainfall gradient is more-or-less constant at 3–6 m/Ma, but that the degree of landscape dissection is differentiated by the relative contribution of chemical versus physical erosion (Chadwick et al., Reference Chadwick, Roering, Heimsath, Levick, Asner and Khomo2013).

The role of chemical versus physical erosion of a land surface is mediated by water, and where relief is steep, it is anticipated that physical erosion would dominate. This is verified in the high Drakensberg which is eroding at a rate of 6 m/Ma (Flemming et al., Reference Fleming, Summerfield, Stone, Fifield and Cresswell1999) while the Gamsberg, which is subject to a much drier regime, is reducing at only 0.4 m/Ma (Cockburn et al., Reference Cockburn, Brown, Summerfield and Seidl2000). The retreat of the Gamsberg escarpment is taking place at about 10 m/Ma while the Drakensberg escarpment is retreating at 50–95 m/Ma. The hyper-arid Namib Desert below the escarpment in Namibia is denuding at 0.5–1.0 m/Ma whilst a wetter period commencing at ~2.8 Ma brought about the incision of prominent canyon features in this landscape (van der Wateren and Dunai, Reference van der Wateren and Dunai2001).

High relief and high erosion rates are typically coupled (Portenga and Bierman, Reference Portenga and Bierman2011) and accordingly ancient landscapes should self-level in the absence of faulting or tectonic uplift that alters the dynamics. The mountains of the southern Cape are enigmatic because they possess high relief and steep slope gradients and yet they are eroding at about 5.2 m/Ma, approximately 100 times slower than regions of the Alps and the Andes ranges with similar relief (Scharf et al., Reference Scharf, Codilean, de Wit, Jansen and Kubik2013). This is attributed to the chemically inert and physically robust rocks that make up the orogen, and accordingly implies that the denudation is not weathering-limited. Although rock mechanics explain the slow erosion rate, Scharf et al. (Reference Scharf, Codilean, de Wit, Jansen and Kubik2013) also report low relief dependence of denudation, as both interfluves and slopes were eroding at the same rate. The implication is that the southern Cape mountain relief is not self-levelling, but that all aspects of the relief are eroding at the same rate, and accordingly that there has been tectonic and topographic stability through the Cenozoic (Scharf et al., Reference Scharf, Codilean, de Wit, Jansen and Kubik2013). In contrast to this, Bierman et al. (Reference Bierman, Coppersmith, Hanson, Neveling, Portenga and Rood2014) report that pediment surfaces in the same region are eroding at a rate that is approximately an order of magnitude lower than the ‘landscape as a whole’. The maintenance of steep relief is attributed to a continental margin effect in which uplift is of isostatic rather than tectonic origin. As the landscape loses mass through erosion, it rises, and in the case of the southern Cape mountains this slow isostatic uplift is sufficient to maintain and even slowly enhance the characteristic relief of the region (Bierman et al., Reference Bierman, Coppersmith, Hanson, Neveling, Portenga and Rood2014). Comparing the long-term incision rate of the Sundays River over the last 4 Ma (16.1 m/Ma) with a 4.26 Ma elevated marine terrace near Durban, yields an isostatic uplift rate of 9.4 m/Ma (Erlanger et al., Reference Erlanger, Granger and Gibbon2012), confirming that isostasy and not tectonic uplift is the dominant mechanism in southern Africa during the Neogene. A common theme in the study of landscape denudation is that denudation rates are highest in the wettest areas. This has also been demonstrated on short time scales (e.g. Tooth et al., Reference Tooth, McCarthy, Brandt, Hancox and Morris2002, Reference Tooth, Brandt, Hancox and McCarthy2004; Rodnight et al., Reference Rodnight, Duller, Tooth and Wintle2005; Keen-Zebert et al., Reference Keen-Zebert, Tooth, Rodnight, Duller, Roberts and Grenfell2013), although these conclusions were reached on the basis of luminescence and radiocarbon dating of resulting sediments rather than on landscape erosional features themselves.

7.3.2 Aggrading deposits

The sedimentary by-product of landscape denudation is transported at different times by colluvial, fluvial and aeolian processes. All of these processes lend themselves to luminescence dating, as the transport process usually presents some degree of sunlight exposure that may be sufficient to re-set the luminescence clock to zero.

One of the earliest applications of luminescence dating in southern Africa was on the colluvial succession at St. Paul’s Mission in KwaZulu-Natal (Wintle et al., Reference Wintle, Li and Botha1993). This was done using the IRSL technique on the fine (<10 μm) sediment fraction, but subsequent comparisons with radiocarbon dates from the same site (Wintle et al., Reference Wintle, Li, Botha and Vogel1995a) demonstrated that colluvial transport was inadequate to fully zero the luminescence signal. A re-analysis of larger sediment grains using TL and IRSL dates, which have different bleaching characteristics, as well as radiocarbon dates, concluded that the IRSL method yielded more accurate ages. The results suggested that episodic deposition took place 100 kyr, 56–50 kyr, post-45 kyr, and within the last 2 kyr (Wintle et al., 1995b). The St. Paul’s study illustrates the complexity that besets application of luminescence dating as a result of inadequate bleaching of the luminescence signal during sediment deposition, and the necessity to independently validate (here using radiocarbon dating) if results reflect the true age of the target event. Since this study, OSL methods that make use of a very rapidly bleaching luminescence signal have been developed, enabling colluvial processes to be more accurately dated. For example, Lyons et al. (Reference Lyons, Tooth and Duller2013) dated colluvial and donga (gully) formation processes between 22 and 1.6 kyr BP and concluded that the most recent period of donga formation (since 890 years BP) is likely to have been driven by climate change and not poor land management practices.

Although not strictly aggrading deposits, different river terraces in the Sundays River, Eastern Cape, have been dated to between 4.0 and 0.3 Ma using cosmogenic dating (Erlanger et al., Reference Erlanger, Granger and Gibbon2012). The same technique suggests that the oldest alluvial deposits in the Vaal River (Rietputs Formation) are approximately 1.57 Ma (Gibbon et al., Reference Gibbon, Granger, Kuman and Partridge2009). This deposit contains Early Stone Age (Acheulean) hand-axes that are presumed to have concentrated through alluvial and colluvial processes, and as a secondary deposit the age is presumed to be a minimum estimate. Further luminescence dating of alluvial deposits has focused on extreme flood event histories (Boshoff et al., Reference Boshoff, Kovacs, Van Bladeren and Zawada1993; Benito et al., Reference Benito, Thorndycraft, Rico, Sánchez-Moya, Sopeña, Botero, Machado, Davis and Pérez-González2011; Grodek et al., Reference Grodek, Benito, Botero, Jacoby, Porat, Haviv, Cloete and Enzel2013), or the geomorphological controls on landscape evolution through alluvial processes (Tooth et al., Reference Tooth, McCarthy, Brandt, Hancox and Morris2002, Reference Tooth, Brandt, Hancox and McCarthy2004, Reference Tooth, Hancox, Brandt, McCarthy, Jacobs and Woodborne2013; Bourke et al., Reference Bourke, Child and Stokes2003; Eitel et al., Reference Eitel, Kadereit, Blümel, Hüser and Kromer2005, Reference Eitel, Kadereit, Blümel, Hüser, Lomax and Hilgers2006; Rodnight et al., Reference Rodnight, Duller, Tooth and Wintle2005; Marren et al., Reference Marren, McCarthy, Tooth, Brandt, Stacey, Leong and Spottiswoode2006; Srivastava et al., Reference Srivastava, Brook, Marais, Morthekai and Singhvi2006).

When rivers reach the sea, the fate of the sediment load is determined by ocean current and wave energetics. The feature of interest here is the almost ubiquitous terrestrial sand dunes that are found around the coast of southern Africa. The deposition of these dunes involves a period of aeolian transport that almost guarantees that the sand grains were exposed to sufficient sunlight at the time of deposition, and that the luminescence signal is completely zeroed. Accordingly, dunes are easy to date using luminescence techniques and their chronological patterns reflect climatic as well as sea-level changes through time. Luminescence dating of coastal dunes also established the ages of intra-formational human footprints preserved in the Nahoon dunes (124 kyr BP; Jacobs and Roberts, Reference 117Jacobs and Roberts2009) and Langebaan dunes (~120 kyr BP; Roberts, Reference Roberts2008), or the similar aged anatomically modern femur found in the dunes near Blind River (Wang et al., Reference Wang, Tobias, Roberts and Jacobs2008). Most dating of coastal dunes is aimed at determining dune dynamics and formation history, and has been undertaken in Mozambique (Armitage et al., Reference Armitage, Botha, Duller, Wintle, Rebêlo and Momade2006), KwaZulu-Natal (Botha et al., Reference Botha, Bristow, Porat, Duller, Armitage, Roberts, Clarke, Kota, Schoeman, Bristow and Jol2003; Sudan et al., Reference Sudan, Whitmore, Uken and Woodbourn2004; Cawthra et al., Reference Cawthra, Uken and Ovechkina2012), the Southern Cape (Carr et al., Reference Carr, Thomas, Bateman, Meadows and Chase2006, Reference Carr, Bateman and Holmes2007, Reference Carr, Bateman, Roberts, Murray-Wallace, Jacobs and Holmes2010) and Western Cape (Fuchs et al., Reference Fuchs, Kandel, Conard, Walker and Felix‐Henningsen2008; Roberts et al., Reference 119Roberts, Bateman, Murray-Wallace, Carr and Holmes2009).

Alluvial deposits in the Namib Desert have been dated at several locations to <130 kyr BP but most date within the last 50 kyr and are related to Pleistocene climate changes in precipitation (Stone and Thomas, Reference Stone and Thomas2013). The residence time of sand in the Namib Desert is estimated at 1 Ma based on cosmogenic dating (Vermeesch et al., Reference Vermeesch, Fenton, Kober, Wiggs, Bristow and Xu2010), but the dynamics of sand dune migration yields much younger luminescence dates for specific sand bodies. The age-frequency of sand dunes was thought to indicate ubiquitous climate changes through time (e.g. Stokes et al., Reference Stokes, Thomas and Washington1997) but this was an artefact of inadequate sample densities, and age gaps have subsequently been filled (Stone and Thomas, Reference Stone and Thomas2013; Thomas and Burrough, Reference Thomas and Burrough2013). Detailed age mapping within dunes, combined with structural evidence from ground penetrating radar, demonstrates that dune ages are controlled by the pace and direction of their migration (Stone and Thomas, Reference Stone and Thomas2013; Thomas and Burrough, Reference Thomas and Burrough2013). The maximum age is determined by the reconstitution time (in which 100% of the sand body has moved as part of the dune migration). Luminescence dating is now questioning the underlying assumption that dune turnover reflects landscape-wide climate forcing. Examining lacustrine (Burrough and Thomas, Reference Burrough and Thomas2008) and fluvial deposits within dune systems (Stone and Thomas, Reference Stone and Thomas2013), that are dated using the same approach as that used for dunes, may be a more direct way to identify the palaeo-moisture regime.

7.3.3 Palaeoanthropology and archaeological deposits

The southern African landscape preserves some of the richest hominid fossil-bearing deposits in the world, and in addition a record of human development is preserved in the Early Stone Age (ESA), Middle Stone Age (MSA), and Later Stone Age (LSA) archaeological records. Initial dating of hominid fossil sites relied on comparisons between the associated fauna, and the well-dated faunal assemblages from east Africa (e.g. Vrba, Reference Vrba1975). The first attempt to obtain a more direct date for the Sterkfontein fossils was the application of ESR dating that yielded a date of 2.1 Ma (Schwarcz et al., Reference Schwarcz, Grün and Tobias1994). The prominence of the fossil StW 573 (Littlefoot) discovered at Sterkfontein Cave demanded better dating, and Partridge et al. (Reference Partridge, Granger, Caffee and Clarke2003) generated an age of 4.02 Ma using cosmogenic dating constrained by evidence for palaeomagnetic reversals at the site. Using the same approach, the age of Australopithecis sediba from the Melapa site was assigned an age of 1.95–1.78 Ma (Dirks et al., Reference Dirks, Kibii, Kuhn, Steininger, Churchill, Kramers, Pickering, Farber, Mériaux, Herries, King and Berger2010) and was later revised to 1.977 Ma (Pickering et al., Reference Pickering, Dirks, Jinnah, de Ruiter, Churchill, Herries, Woodhead, Hellstrom and Berger2011). Subsequently, the application of U/Pb dating yielded an age of 2.2 Ma for the flowstone associated with the Littlefoot fossil (Pickering and Kramers, Reference Pickering and Kramers2010). Some reconciliation between the cosmogenic and U/Pb dates can be reached by using an updated half-life for ¹⁰Be (Partridge, Reference Partridge2005), but the two dating approaches still yield different results.

Despite the abundance of Acheulian hand-axes in the ESA, there are very few locations at which they have been dated. The assemblage in the Vaal River gravels predates 1.57 Ma based on cosmogenic dating (Gibbon et al., Reference Gibbon, Granger, Kuman and Partridge2009). The only site at which the ESA is directly dated is at Wonderwerk Cave where palaeomagnetic dating, luminescence and cosmogenic dating yielded an age range of 1.96–0.78 Ma (Chazan et al., Reference Chazan, Ron, Matmon, Porat, Goldberg, Yates, Avery, Sumner and Horwitz2008). The Fauresmith Industry, considered to be transitional between the ESA and MSA, is dated to 542–464 kyr BP using luminescence and combined U-series/ESR dating at Wonderwerk Cave (Porat et al., Reference Porat, Chazan, Grün, Aubert, Eisenmann and Horwitz2010). This appears to predate the late Acheulian assemblage from Duinefontein II, which has IRSL dates of 292 kyr and 265 kyr BP (Feathers, Reference Feathers2002). These apparently inverted dates challenge the perception that the ESA transition to the MSA was a temporally coherent phenomenon, while also raising questions regarding the technological classifications that have been applied in assigning stone assemblages to different industries.

Problems with early attempts to date MSA sites in southern Africa arose because the deposits were older than the 50 kyr limit of radiocarbon dating. One of the first MSA stone tool technological complexes identified was the Howiesons Poort (named after the type-site) which yielded a finite radiocarbon date of 29,400±675 ¹⁴C BP at Diepkloof Cave (Parkington and Poggenpoel, Reference Parkington, Poggenpoel, Parkington and Hall1987). This date was interpreted literally in the absence of any other dated assemblages from this time period (Parkington, Reference 118Parkington and Mellars1990). In contrast, the equivalent assemblage from Border Cave was dated by extrapolation of deposition rates of the uppermost layers at the site that were radiocarbon dated. This yielded an age estimate of 95 kyr for the Howiesons Poort and 195 kyr for the base of the MSA deposits (Butzer et al., Reference Butzer, Beaumont and Vogel1978). The date was treated with scepticism because of the implied age of 115–90 kyr BP for the human remains recovered at the site. However, the application of isoleucine epimerisation dating to the Howiesons Poort industry at Border Cave, Apollo 11 Cave, and Boomplaas Cave placed the industry in the age range 80–56 kyr BP with a most likely age of 66±5.0 kyr BP (Miller et al., Reference Miller, Beaumont, Deacon, Brooks, Hare and Jull1999). Further application of U/Th dating at Boomplaas and Klasies River Mouth placed the Howiesons Poort at 70–60 kyr BP (Vogel, Reference Vogel, Tobias, Raath, Moggi-Cecchi and Doyle2001).

The emerging antiquity of the Howiesons Poort and the MSA from southern Africa more generally has important global implications. In particular the recovery of anatomically modern human remains from Border Cave and Klasies River Mouth as well as subsequent discoveries of evidence for modern cognitive behaviour at Blombos Cave (Henshilwood et al., Reference Henshilwood, d’Errico, Yates, Jacobs, Tribolo, Duller, Mercier, Sealy, Valladas, Watts and Wintle2002) highlighted the significance of this geochronometric record. Subsequently, Howiesons Poort sediment at Rose Cottage Cave was dated to 70–60 kyr BP (Pienaar et al., Reference Pienaar, Woodborne and Wadley2008) and burned lithics from the same site to 60.4–56.3 kyr BP (Valladas et al., Reference Valladas, Wadley, Mercier, Froget, Tribolo, Reyss and Joron2005). At Klasies River Mouth, Howiesons Poort sediments were dated to 60–50 kyr BP (Feathers, Reference Feathers2002) and 56±3.0 kyr BP from the same site using burned quartz samples (Tribolo et al., Reference Tribolo, Mercier and Valladas2005). At Diepkloof Shelter, luminescence dating yields Howiesons Poort ages of 65–55 kyr BP (Tribolo et al., Reference Tribolo, Mercier and Valladas2005). Sibudu Cave yielded luminescence dates of 61.1±1.5 kyr BP (Wadley and Jacobs, Reference Wadley and Jacobs2006) and Blombos Cave yielded minimum ages of 70.1±1.9 kyr BP (Jacobs et al., Reference Jacobs, Wintle and Duller2003a) and 68.8±3.0 kyr BP (Jacobs et al., Reference Jacobs, Duller and Wintle2003b).

While the luminescence dating of the Howiesons Poort appears to be converging, there remains a lack of precision that results in a chronological ‘haze’ (Jacobs et al., Reference Jacobs, Roberts, Galbraith, Deacon, Grün, Mackay, Mitchell, Vogelsang and Wadley2008). Part of this is attributed to the different assumptions and laboratory protocols implemented by different practitioners, and accordingly systematic re-dating of sites containing Howiesons Poort and the preceding Still Bay industries was conducted at a sub-continental scale. This indicates that the Howiesons Poort is 64.8–59.5 kyr BP and this is clearly distinguished from the Still Bay of 71.9–71.0 kyr BP (Jacobs et al., Reference Jacobs, Roberts, Galbraith, Deacon, Grün, Mackay, Mitchell, Vogelsang and Wadley2008). To a large degree, this result is corroborated by ESR and radiocarbon dates from Border Cave (Grün and Beaumont, Reference Grün and Beaumont2001; Bird et al., Reference 114Bird, Fifield, Santos, Beaumont, Zhou, di Tada and Hausladen2003).

The transition from the MSA to the LSA in southern Africa is characterised by stone tool assemblages that are poorly defined and are assigned to the umbrella Middle Stone Age III (MSA III) and the Early Later Stone Age (ELSA). Although such deposits occur at many of the stratified MSA sites, spatial inter-comparisons are hampered by the lack of typologically diagnostic stone tool assemblages. The date brackets for this transition is ~55–37 kyr BP at Border Cave (Grün and Beaumont, Reference Grün and Beaumont2001; Bird et al., Reference 114Bird, Fifield, Santos, Beaumont, Zhou, di Tada and Hausladen2003), and ~52–26 kyr BP at Sibudu Cave (Wadley and Jacobs, Reference Wadley and Jacobs2004).

The LSA in southern Africa has been recovered from many well-stratified sites (including many of the MSA sites mentioned above), and better preservation of appropriate organics in these deposits has allowed accurate and precise application of radiocarbon dating. A key element here has been the improvement in the calibration dataset. In 1998 the calibration of radiocarbon dates up to 10.3 kyr BP was based on a robust tree ring chronology, and alternative datasets extended the calibration to 24 kyr with increasing levels of uncertainty (Stuiver et al., Reference Stuiver, Reimer, Bard, Beck, Burr, Hughen, Kromer, McCormac, van der Plicht and Spurk1998). Dates older than 24 kyr could not be calibrated. By 2004 the calibration limit was pushed back to 26 kyr BP (Reimer et al., Reference Reimer, Baillie, Bard, Bayliss, Beck, Bertrand, Blackwell, Buck, Burr, Cutler, Damon, Edwards, Fairbanks, Friedrich, Guilderson, Hogg, Hughen, Kromer, McCormac, Manning, Ramsey, Reimer, Remmele, Southon, Stuiver, Talamo, Taylor, van der Plicht and Weyhenmeyer2004), but by 2013 reasonable calibrations became possible for the full 50 kyr range of the technique (Reimer et al., Reference Reimer, Bard, Bayliss, Beck, Blackwell, Ramsey, Buck, Cheng, Edwards, Friedrich, Grootes, Guilderson, Haflidason, Hajdas, Hatté, Heaton, Hoffmann, Hogg, Hughen, Kaiser, Kromer, Manning, Niu, Reimer, Richards, Scott, Southon, Staff, Turney and van der Plicht2013). A southern hemisphere radiocarbon reservoir offset demanded a specific calibration dataset (Hogg et al., Reference Hogg, Hua, Blackwell, Niu, Buck, Guilderson, Heaton, Palmer, Reimer, Reimer, Turney and Zimmerman2013), and this offset was factored in to the early Pretoria calibration (Vogel et al., Reference Vogel, Fuls, Visser and Becker1993) that was employed by most radiocarbon users in southern Africa. Many of the dates from the earlier part of the LSA that were inadequately calibrated could later be placed into an absolute dating reference when the calibration datasets were extended. Together this led to high spatial and temporal resolution in the LSA for southern Africa, and the chronology for the last 30 kyr is relatively robust.

The Iron Age in southern Africa is the part of the archaeological heritage that is most relevant to the movement and economies of the peoples that eventually led to the current human geography of the region. It has powerful political ramifications in the identities, cultures and land rites in the region. Dating has mainly been accomplished using radiocarbon dating with exceptional precision. A summary of Iron Age dating is provided by Huffman (Reference Huffman2007) where the association between traditions of pottery design and motifs is correlated with age. The use of pottery motifs as an age-indicator in the Iron Age eclipses radiocarbon dating because of its ease of application, but it is still necessary to acknowledge the radiocarbon dates on which the pottery chronology is based. This presents problems in the dating of certain sites such as furnaces where ceramics may not be recovered and where the problem of old wood may influence the use of radiocarbon dating. Luminescence dating may offer an alternative dating technique provided the dose rate for the site can be sufficiently constrained, and attempts to do this have been undertaken, but have not yet been published.

Within the Iron Age, the radiocarbon dates allow sufficient precision such that they can be analysed in their own right. Huffman (Reference Huffman2010) considered the dating evidence for the burning of cattle byres (kraals) and, combined with ethnographic evidence, showed that such practices are associated with severe droughts, identifying 13 episodes of kraal burning across the subcontinent over the last 2000 years. These were correlated with El Niño events from southern Ecuador to suggest a common climatic origin. Although the Iron Age dates from southern Africa can be analysed in this way when aggregated, there are often problems with inadequate dating within individual sites. Examples of this are highlighted below.

7.3.4 Palaeoclimate

There are a number of depositional sequences in southern Africa that yield evidence of past climate. Some of these, such as speleothems, have excellent chronological control. The most prominent speleothem records are the 25 kyr record from Cold Air Cave (Holmgren et al., Reference Holmgren, Lauritzen and Possnert1994, Reference Holmgren, Lee-Thorp, Cooper, Lundblad, Partridge, Scott, Sithaldeen, Talma and Tyson2003), the 50 kyr record from Cango Cave (Talma and Vogel, Reference Talma and Vogel1992) and the 90–50 kyr Pinnacle Point record (Bar-Matthews et al., Reference Bar-Matthews, Marean, Jacobs, Karkanas, Fisher, Herries, Brown, Williams, Bernatchez, Ayalon and Nilssen2010). In each case, speleothems have been dated using U-series disequilibrium dating, and this technique does not commonly yield age inversions unlike many sedimentary sequences dated using radiocarbon. The 20 kyr Wonderkrater spring mound record yielded a number of radiocarbon age inversions (Scott et al., Reference Scott, Bryant, Cook and Naysmith2003), which is likely to be a function of the carbon systematics on the site. In other applications, such as the rock hyrax midden palaeoclimate records that extend back to 20 kyr BP (Chase et al., Reference 115Chase, Meadows, Scott, Thomas, Marais, Sealy and Reimer2009, Reference Chase, Meadows, Carr and Reimer2010, Reference Chase, Quick, Meadows, Scott, Thomas and Reimer2011), radiocarbon provides a coherent chronology. Dating of the Tswaing impact crater record has also been improved by a radiocarbon chronology that extends to 50 kyr BP, but this only covers the deposit to about 18 m depth in a total sequence of 80 m, and the basal date is fixed on a single fission-track date of 220 kyr BP (Kristen et al., Reference Kristen, Fuhrmann, Thorpe, Röhl, Wilkes and Oberhänsli2007). The original age model used the basal date and radiocarbon dates to establish an accumulation rate, and relied on orbital tuning to refine the chronology (Partridge et al., Reference Partridge, Demenocal, Lorentz, Paiker and Vogel1997). There is clearly scope for improved dating resolution in the 50–220 kyr BP range.

7.4.1 Interpretation of the dating context

Despite the maturity of some dating techniques, anomalous ages still emerge either within sequences that are dated by a single method, or between results for a single ‘event’ obtained using different methods. An example is a radiocarbon date of 28,880±170 ¹⁴C years BP on a notched bone recovered at Sibudu Cave, but it is attributed to an age of 61.3±2.0 kyr BP based on the OSL method (Wadley and Jacobs, Reference Wadley and Jacobs2006). Another radiocarbon date from the same site of 34,300±2000 ¹⁴C years BP on charcoal is rejected for the same reason, whilst another of 24,200±290 ¹⁴C years BP is rejected because of an inverted age sequence (Wadley and Jacobs, Reference Wadley and Jacobs2004). These examples illustrate the problem of presuming which dates are valid and which are invalid. The Sibudu Cave authors presume that there is an error in the radiocarbon analysis, arguing that the standard pre-treatment of samples may not adequately remove contaminants that accumulate in charcoal (see Bird et al. (Reference Bird, Levchenko, Ascough, Meredith, Wurster, Williams, Tilston, Snape and Apperley2014) for discussion of this protocol). They claim that ‘Until multiple radiocarbon ages have been obtained for the same layer from samples pre-treated in an appropriate way, as was the case at Border Cave (Bird et al., Reference 114Bird, Fifield, Santos, Beaumont, Zhou, di Tada and Hausladen2003), meaningful comparisons cannot be made to other dating techniques, such as OSL dating’ (Wadley and Jacobs, Reference Wadley and Jacobs2006, p5). The cited ‘case at Border Cave’ is directly comparable with the radiocarbon dates at Sibudu Cave. Radiocarbon dates produced for Border Cave by the same laboratory that provided the rejected Sibudu dates and using the same pre-treatment protocols, fell in a range 37.7±0.6 kyr to 39.8±0.6 ¹⁴C kyr BP, while the specialised pre-treatment that is presumed to distinguish the Border Cave results yields a date of 38.5 ¹⁴C kyr BP (Bird et al., Reference 114Bird, Fifield, Santos, Beaumont, Zhou, di Tada and Hausladen2003). Thus, pre-treatment of the samples may be rejected as a significant source of error. The rejected Sibudu date of 34,300 ¹⁴C years BP calibrates to 34–39 kyr BP (rounded to the nearest thousand years, using the calibration of Hogg et al. (Reference Hogg, Hua, Blackwell, Niu, Buck, Guilderson, Heaton, Palmer, Reimer, Reimer, Turney and Zimmerman2013)). This is the age of the immediately overlying sediments (dated by OSL) at Sibudu Cave, and it is very likely that the charcoal that yielded the rejected date may be derived from the younger occupation of the site.

Whereas the Sibudu example centres on the question of rejecting dates, the provenance of dates that are accepted is also a critical part of the chronological framework. Recently, Chirikure et al. (Reference Chirikure, Pollard, Manyanga and Bandama2013) proposed a revision of the chronology for Great Zimbabwe using a Bayesian approach to the existing dataset of radiocarbon dates. They make use of the 2004 southern hemisphere calibration dataset (McCormac et al., Reference McCormac, Hogg, Blackwell, Buck, Higham and Reimer2004) and presume that this is better than the calibration used by Huffman and Vogel (Reference Huffman and Vogel1991) which involved using the northern hemisphere calibration curve in a southern hemisphere application. Chirikure et al. (Reference Chirikure, Pollard, Manyanga and Bandama2013) argue from their re-analysis that the emergence of Great Zimbabwe as a regional capital took place around AD 1225, coeval with Mapungubwe. A key element here is identifying the correct stratigraphic position of stone wall structures that is indicative of the development of the site within the dating sequence. While Chikure et al. (Reference Chirikure, Pollard, Manyanga and Bandama2013) place this interface between the stratigraphic units called ‘Period II’ and ‘Period III’, Huffman (Reference Huffman2015) places it between ‘Period IVa’ and ‘Period IVb’. The problem that leads to this discrepancy in the Iron Age chronology of southern Africa is not the result of ‘dating hygiene’ as suggested by Chirikure et al. (Reference Chirikure, Pollard, Manyanga and Bandama2013), but rather a problem of archaeological interpretation.

These examples suggest that greater confidence should be placed on the veracity of measured dates, provided that the context of the sample is fully understood. However, it is possible that contradictory dates will arise when the dating technique links to the target event via a set of assumptions that are not necessarily applicable in the given context.

7.4.2 Insight into the suitability of the dating technique to the target event

Early applications of radiocarbon dating to pedogenic calcretes were compared with the age results obtained with TL and U/Th dating (e.g. Geyh and Eitel, Reference 116Geyh and Eitel1998). This analysis emphasised that each method was dating a different formational event, but there has commonly been a literal and uncritical interpretation of radiocarbon ages derived from calcrete deposits. Many of these ages have been overturned with the advent of luminescence techniques and this has led to after-the-fact accommodation of the problems associated with radiocarbon dating of pedogenic carbonates. In this instance it is not possible to ascertain if radiocarbon dates are accurate unless an independent ageing technique is applied, because under ideal circumstances the radiocarbon dates may be indeed correct (Geyh and Eitel, Reference 116Geyh and Eitel1998). The consistency of radiocarbon dates within a sequence, and the consistency of associated data, is often used as a defence of the radiocarbon method on pedogenic carbonates. The dating of heuweltjies (round pedological features up to 5 m in diameter) in the western Cape has yielded systematic age/depth relationships, and the associated isotopic values of the carbonate are believed to retain a climatic signal, implying minimal post-depositional alteration of the carbonate matrix (Midgley et al., Reference Midgley, Harris, Hesse and Swift2002; Potts et al., Reference Potts, Midgley and Harris2009). In such instances, both radiocarbon ages and isotopic values should be treated with caution until independent data confirm these. Radiocarbon dating of glacial and periglacial features in the Drakensberg to the last glacial period (Mills and Grab, Reference Mills and Grab2005) has a similar criticism. These authors argue why the radiocarbon method is dating the target event, but the range of influences on soil carbon that might nullify the result cannot be disproven. In this instance, the results should be verified by alternative dating techniques such as cosmogenic ‘burial dating’.

In any circumstance in which multiple dating techniques are applied, there exists a possibility that discrepancies will arise in the results. These may be because of systematic errors inherent in the technique, or that the techniques are dating different events. In the foregoing discussion, radiocarbon dating was demonstrated to be problematic in some applications, but the radiocarbon dating community has regular blind inter-laboratory performance comparisons using known-age samples, and there is a high level of consistency of results between laboratories (Scott et al., Reference Scott, Bryant, Cook and Naysmith2003). However, this confidence does not extend to all dating techniques (see Jacobs et al., Reference Jacobs, Roberts, Galbraith, Deacon, Grün, Mackay, Mitchell, Vogelsang and Wadley2008) and thus multiple dating techniques can be applied to situations in which radiocarbon is known to yield a good chronology. For example, at Rose Cottage Cave both radiocarbon and luminescence dating yielded very similar results (Pienaar et al., Reference Pienaar, Woodborne and Wadley2008). Whereas all the radiocarbon dates used in that study provided a coherent increase in age with depth, it was the luminescence dates that yielded age inversions. Examples like this raise queries about the assumptions that are made in different dating techniques, such as the effects of changing radiation dose rates used in dosimetry-based techniques like luminescence and ESR dating.

7.4.3 Reproducibility, consistency, and a body of evidence

Identifying erroneous dates usually only emerges from age inversions, or when different techniques produce different dates on the same target event. Post hoc accommodation is the most common means of rejecting erroneous dates, but it relies on the overwhelming evidence from other consistent dates. For example, the large number of luminescence dates for dunes in the Namib and Kalahari deserts (Stone and Thomas, Reference Stone and Thomas2013; Thomas and Burrough, Reference Thomas and Burrough2013) led to a paradigm shift in the use of dunes as a palaeoclimate proxy. Similarly, a total of 71 ESR dates was used to provide the chronology of different MSA layers at Border Cave (Grün and Beaumont, Reference Grün and Beaumont2001; Millard, Reference Millard2006). This is in contrast to the dating of highly significant sites such as Mapungubwe and Great Zimbabwe. Mapungubwe was dated with only 24 radiocarbon dates with an additional 7 on gold artefacts from the site (Woodborne et al., Reference Woodborne, Pienaar and Tiley-Nel2009), and Great Zimbabwe has only 21 radiocarbon dates (Chirikure et al., Reference Chirikure, Pollard, Manyanga and Bandama2013). Thus, the emerging chronology requires many more dates in order to reject inappropriate interpretations. While reproducibility and consistency in age determinations within a sequence, and between the same context in different locations, can provide grounds for the rejection of outliers, this approach is based on the precision of the body of evidence, and rejecting of dates infers knowledge about the accuracy (inaccuracy) of the rejected date. It is more appropriate to flag these dates for closer scrutiny in order to identify independent reasons why the date is deemed inaccurate.