Worldwide, the millennia preceding and encompassing the Last Glacial Maximum (LGM), defined as the phase of maximum global ice volume during the opening millennia of Marine Isotope Stage (MIS) 2 from approximately 24-18 kcal BP (P. U. Clark et al., 2009), coincide with significant shifts in early human adaptive strategies and demography (Chan et al., 2015; Clarkson et al., 2018; Kuhn & Elston, 2002; Soffer & Gamble, 1990a, 1990b). Fluctuating atmospheric and oceanic circulations, a global temperature depression of 6º C, and extensive glaciation transformed environments and resource distribution across the globe (P. U. Clark et al., 2009; P. U. Clark & Mix, 2002; Seltzer et al., 2021).
A nearly synchronous global shift to microlithic stone tool technology, beginning amidst the foreboding climatic oscillations of MIS 3, also reached its full potential during the LGM (Kuhn & Elston, 2002). Microlithic assemblages were diverse, like the environments they developed within. The Solutrean technocomplex of southwest Europe (~25 ka) blended macrolithic, bifacially flaked points with microliths and bladelets (Straus, 2015, 2016). Late Gravettian foragers used backed bladelets alongside regionally diverse points and knives (Polanská et al., 2021; Tomasso et al., 2018). Siberian and Japanese microblade assemblages also incorporated a range of larger burins, endscrapers, and blades (Doelman, 2008; Goebel, 2002; Iwase, 2016; Otsuka, 2017). In southern Africa, microlithization culminated in the end of the Middle to Later Stone Age (MSA/LSA) transition when the unmodified bladelets of the Robberg technocomplex appeared alongside the LGM ~24 ka (Lombard et al., 2012).
Initially, archaeologists doubted that microlithization was a development internal to southern Africa. Goodwin and Van Riet Lowe (1929, p. 149) tentatively suggested that microlithic assemblages arrived via population incursions from further north. Sampson (1974) modified this idea, asserting that microlithic industries had origins in southern Zambia and Rhodesia (Zimbabwe). J.D. Clark (1959) and Humphreys (1972) were among others advocating for migration theories. Models incorporating climate became more popular in the second half of the twentieth century: J. D. Clark (1974) and Phillipson (1976, 1977) suggested that microlithic technology evolved independently in many areas, taking forms most advantageous to the local environment. H.J. Deacon (1976) proposed that microlithic techniques arrived though diffusion and were then modified to suit environmental conditions. Mitchell’s (1988) hypothesis, specific to the Maloti-Drakensberg mountains of highland Lesotho, implicated the LGM as an impetus for microlithization, positing that microlithic technologies were adaptive in “time-stressed” environments (Torrence, 1983). By the end of the twentieth century, “gradualist” theories rejected a sudden population replacement or migration entirely. Built on sites with long occupation histories like Sehonghong, Umhlatuzana, and Rose Cottage Cave, these hypotheses recognized intrasite continuities and subtle temporal trends (A. M. B. Clark, 1997; J. Kaplan, 1990; Mitchell, 1994).
However, variations of the original population replacement model live on. Most recently, Bousman and Brink (2018) proposed that the MSA/LSA transition actually consisted of two transitions centered around demographic events at ~43 and ~25 ka. The earlier – from the Middle Stone Age (MSA) to the Early Later Stone Age (ELSA) – was first expressed at Border Cave in KwaZulu-Natal (d’Errico et al., 2012; Villa et al., 2012). In this context, “ELSA” encompasses East and southern African lithic assemblages dating to between ~45 and 20 ka. These often feature bipolar cores, scaled pieces, and incipient microlithization and bladelet production, but sometimes macrolithic components or MSA-type formal tools (Bousman & Brink, 2018; Lombard et al., 2022). ELSA groups then slowly brought this technology westwards over a period of 20,000 years. A subsequent transition – from the ELSA to the Robberg – was more rapid, spurred by a population bifurcation event ~25 ka and centered on the Southern Cape and uKhahlamba-Drakensberg Escarpment (Behar et al., 2012; Bousman & Brink, 2018). Bousman and Brink’s (2018) model inherently challenges more recent gradualist interpretations and harkens back to the migration and replacement models of the early twentieth century.
In this paper, we test current models of the MSA/LSA transition in southern Africa against two lithic assemblages dated to the onset of the LGM at Melikane Rockshelter, Lesotho: a fully developed Robberg assemblage and its incipient, developing predecessor. Significant intrasite continuities, both between the two assemblages and throughout the site’s 80,000 year-old sequence more broadly, challenge Bousman and Brink’s (2018) rapid replacement hypothesis. Instead, we suggest that the components of the Robberg flaking system developed slowly across a wider region. Episodic population connectivity, either as a risk-reduction mechanism or due to limitations imposed by LGM environments, encouraged the creation of local flaking systems and their subsequent coalescence and diffusion. The flexibility of bladelet technology contributed to its continued refinement across southern Africa and adoption in a variety of environments, not all of them marginal. However, at Melikane in particular, foragers emphasized bladelet production as an available and adequate means to cope with subsistence risk. Thus, the subcontinent-wide fluorescence of the Robberg may have been the product of sporadic information exchange and the extraordinary versatility of bladelet technology, whilst its adoption at Melikane may have been accelerated by the effects of the LGM on a vulnerable highland environment.
Background
Site Description and Excavation History
Melikane Rockshelter is located in the Qacha’s Nek District of eastern Lesotho on the Melikane River, a tributary of the larger Senqu (Orange) (Fig. 1). It faces northeast at an altitude of ~1860 meters and measures 44 meters long and 21 meters deep, with an average roof height of 7.7 meters (Carter, 1978) (Fig. 2). The shelter is one of countless that have formed within the erosive sandstones of the Clarens Formation just beneath its interface with the basalts of the overlying Drakensberg Formation. Myriad high-quality raw materials are available in this landscape, particularly near the contacts of the region’s sedimentary and volcanic lithologies. These include cryptocrystalline silicates (CCS), partially metamorphosed sandstones, fine-grained hornfels, and durable, coarse-grained volcanic dyke material (Carter, 1978; Pazan et al., 2022; Stewart et al., 2012, 2016).
Melikane was first excavated by Patrick Carter in 1974. Carter removed 36 m3 of sediment, distinguishing major layers by perceived depositional differences, but nonetheless excavating in 10 cm spits that crosscut the site’s visible stratigraphy. He identified 7 layers, originally dating Layer 1 to ~1500 BP and Layer 5 to ~43.2 kcal BP, with the site’s oldest levels falling beyond the radiocarbon boundary (Carter, 1978). In 2007, Carter’s trench was reopened to obtain samples for optically stimulated luminescence (OSL) dating (Jacobs et al., 2008). The following year, the ‘Adaptations to Marginal Environments in the Middle Stone Age’ (AMEMSA) project led by B.A. Stewart and G. Dewar opened a new 2 x 3 m trench 1 m east of Carter’s, ultimately identifying 30 layers loosely correlating with individual contexts. Excavations concluded in 2009 upon reaching bedrock (Stewart et al., 2012). AMEMSA used a single-context recording system, dividing thicker contexts into 5 cm spits when necessary. Artifacts were sieved using 1.5 mm mesh for the upper contexts (including those described in this paper) and 3 mm mesh for lower ones, where sediment moisture was high (Stewart et al., 2012).
AMEMSA selected archaeological charcoals from secure contexts for accelerator mass spectrometry (AMS) radiocarbon dating, with all samples processed at the Oxford Radiocarbon Accelerator Unit. Samples that were expected to date >25 ka were processed using rigorous acid-base-wet oxidation stepped-combustion (ABOx-SC) pretreatment when possible (Layers 6–14), while younger samples were processed using the standard acid-base-acid (ABA) protocol (Brock et al., 2010; Stewart et al., 2012). Carter’s (1978) original dates were also recalibrated for comparison. Nine single-grain OSL dates were also obtained from Carter’s trench, spanning the entire Melikane sequence. Occupation pulses at Melikane cluster around ~80, 60, 50, 46–41, 27–23, 9, and 3 ka (Stewart et al., 2012, p. 51). This paper is concerned with the occupation pulse dating to 27-23 kcal BP. This corresponds to AMEMSA Layer 5 (contexts 6–8) and Layer 4 (context 5), both falling within Phase II of the Melikane sequence (cf. Pazan et al., 2022, p. 119).
AMEMSA Layer 5 (contexts 6–8) comprises a single archaeological unit divided into three arbitrary spits. Spit 1 (6–8:1) was directly dated by Stewart et al. (2012, tbl. 1) to 24.2-23.6 kcal BP. A sediment sample (MLK 9) from Carter’s eastern section wall corresponding to the interface of Spits 3 (6–8:3) and 2 (6–8:2) produced a statistically equivalent OSL date of 27.1 ± 1.8 ka for the majority (61.8%) of its quartz grains (Stewart et al., 2012). AMEMSA Layer 4 (context 5) directly overlies Layer 5 (contexts 6–8) and is composed of five spits. It did not contain sufficient organic material for a radiocarbon date and attempts to retrieve an OSL date from this layer failed. However, Carter’s original excavations produced three additional radiocarbon dates of 24.2-23.6, 24.3-23.8, and 23.8-23.3 kcal BP (Carter, 1978). Stewart et al. (2012) list these under Carter’s equivalent of AMEMSA Layer 5 (Carter Layer 3), but Vogel et al. (Vogel et al., 1986, p. 1144) assign them to Carter Layers 2/3 – equivalent to the interface of AMEMSA Layers 4/5 – and note that they are associated with a “microlithic (?) early LSA assemblage.” Multiple sedimentological indicators also support a LGM age for AMEMSA Layer 4 (context 5). This level yielded the sequence’s lowest organic carbon and magnetic susceptibility values, a peak in mean particle size, and large volumes of sand and imbricated gravels. As outlined below, all point to colluvial inwash from a sparsely vegetated, increasingly periglacial landscape (Stewart et al., 2012, p. 51). Moreover, some parts of Layer 4 are separated from Layer 5 by tabular sandstone slabs (Carter, 1978; Stewart et al., 2012). Similar to those in Layer OS at nearby Sehonghong Rockshelter, dated 24.4–23.9 kcal BP (Mitchell, 1994; Pargeter et al., 2017), these may derive from frost-shattering at the LGM’s onset.
Environmental context
Today, the mean annual temperature (MAT) around Melikane is ~13º C (S. W. Grab, 1997). The site is located in southern Africa’s summer rainfall zone (SRZ), in which >66% of precipitation falls during the summer months (Roffe et al., 2019). Generally, summers are warm with violent thunderstorms, and winters are cold and dry. Snow can fall at any time of year (S. W. Grab, 1997). By 24 kcal BP, a 5-6º C drop in temperatures (Holmgren et al., 2003; Kulongoski et al., 2004; Partridge et al., 1999; Seltzer et al., 2021; Stute & Talma, 1998) and a shift in precipitation seasonality resulted in glacier and permafrost formation on the cold, south-facing slopes of the Escarpment summit and Lesotho’s adjacent high plateau, evidenced by moraines (S. W. Grab, 1996; Mills & Grab, 2005) and periglacially patterned ground (Bregman & Knight, 2022; S. Grab et al., 2021; Mills & Grab, 2005; Wang & French, 1995).
Lower temperatures during the LGM would have affected the vegetation – and thus faunal resources – available to foragers in the Maloti-Drakensberg Mountains. Today, at higher altitudes, lower growing season temperatures favor plants using a C3 photosynthetic pathway, including alpine grasses, which have little nutritional value except for a short period in the summer (Vogel, 1983). Lower altitudes and warmer temperatures favor perennially nutritious C4 tropical grasses, which currently dominate the Melikane River Valley (Ehleringer et al., 1997). However, phytoliths and soil organic matter (SOM) from the site’s early LGM deposits suggest that C3 grasses heavily dominated the local landscape (Stewart et al., 2016) (Fig. 3).
SOM samples were taken at depth intervals of 10 cm, and include two each from Layer 5 (contexts 6–8, spits 3 and 1) and Layer 4 (context 5, spits 3 and 1). The δ13C values from the SOM in contexts 6–8 in Layer 5 are -23.1 ‰ and -22.5 ‰ in the lower and upper samples, respectively. In the layer above, the lower context 5 sample yielded a value of -23.2 ‰, comparable to the values from contexts 6–8, while the upper context 5 sample is the most negative in the entire Melikane sequence at -24.0 ‰.
Bousman (1991) developed an index specifically for use in the uKhahlamba-Drakensberg Escarpment to predict the proportion of C3 grasses on the landscape based on the results of bulk SOM δ13C values. His index indicates a C3 plant component of ~85% for the bottom of Layer 5 (contexts 6–8, spit 3), rising to ~90% in the top of that layer (contexts 6–8, spit 1) and the bottom of Layer 4 (context 5, spit 3). The top sample in Layer 4 (context 5, spit 1) produced the highest Bousman Index values of the sequence, suggesting an environment with >95% C3 grasses. The C4 grasses that remained were drought-tolerant chloridoids (Stewart et al., 2016; Twiss, 1992). The tree cover density ratio (D/P ratio), defined as the proportion of woody to grassy phytolith morphotypes, plunged to its lowest point in the Melikane sequence in Layer 5 (contexts 6–8) and rose slightly in Layer 4 (context 5) (Stewart et al., 2016). This suggests a relatively barren landscape on the eve of the LGM, with tree cover increasing incrementally over time alongside the expansion of C3 grasses.
Charcoal samples were obtained from the lower of the two layers under consideration – Layer 5 (contexts 6–8). Although charcoal assemblages are inherently biased towards preferred sources of fuel and do not reflect proportions of taxa, they are impotent indicators of the habitats available near the site. Three species – Erica drakensbergensis, Leucosidea sericea, and Olea europea – are well represented in the early MIS 2 assemblage (Stewart et al., 2016). Today, Erica drakensbergensis is found in moist areas, but is also able to withstand dry conditions. It is a hardy, frost resistant taxon that toleratesharsh, cold conditions (Mucina & Rutherford, 2006). Olea europaea (the wild olive) is a highly successful and hardy plant that can survive some frost, but tends to thrive in warmer areas with MATs <15ºC (Mucina & Rutherford, 2006). Yet, olives can still germinate in cold environments provided that there is a sufficient difference between winter and spring temperatures (Orlandi et al., 2005). Protea sp. is more indicative of drier, open scrublands than the other taxa, which are common riverine species. Like Leucosidea sericea, it is also frost-resistant. Relative to the charcoals from Melikane’s pre-MIS 2 levels, there is narrowing of woody species in Layer 5. Notably absent are Buddleja salviifolia and Rhamnus sp., both of which occur throughout the earlier deposits. While Rhamnus sp. is resistant to frost, Buddleja salviifolia, which is a parallel scrub community to Leucosidea sericea, exists at lower altitudes than the latter and is less cold-hardy (Killick, 1978).
Overall, the analyses of phytoliths, SOM δ13C and archaeological charcoals from Layers 5 and 4 at Melikane indicate that the site was surrounded by a cool, open C3 grassland, which became even colder after 24 kcal BP. Carter (1976) hypothesized that during particularly harsh periods such as the LGM, the highlands may have been abandoned altogether. However, the landscape was clearly still productive enough to support populations at the onset of glacial conditions. Future attempts to date Layer 4 (context 5) will hopefully clarify the timing of Melikane’s abandonment during the LGM, but paleoenvironmental indicators from the layer are consistent with the coldest conditions experienced by humans in the Lesotho Highlands.