While our understanding of variability in hominin behavior in the African Early Stone Age (ESA) has increased substantially over the last two decades, much of this information derives from the eastern African rift system where well-constrained geochronologies and tightly associated material remains yield detailed data on hominin adaptive responses to evolving landscapes through the Plio-Pleistocene (Kingston, 2007; Archer et al., 2014; de la Torre et al., 2014; Presnyakova et al., 2018; Braun et al., 2019; Reeves et al., 2021). Comparative contexts in southern Africa are largely confined to non-occupational cave deposits (Herries and Shaw, 2011; Granger et al., 2015; Kramers and Dirks, 2017; Pickering et al., 2019), whereas well-contextualized cave occupations are largely confined to the last ∼200 kyr (Jacobs et al., 2008; Marean, 2010; Tribolo et al., 2013; Wilkins et al., 2021).

The paucity of southern African ESA sites with well-constrained evidence for hominin occupation is particularly evident in the period between ∼1.8 Ma and 0.3 ka, which is conventionally associated with the Acheulean techno-complex (Herries, 2011; Braun et al., 2016; Duke et al., 2021). The Acheulean is traditionally recognized through the presence of a characteristic tool type known as Large Cutting Tools (LCTs). Early assemblages with LCTs are thought to represent the onset and proliferation of several behavioral innovations, including the expansion of foraging ranges and raw material procurement, increased depth of planning, expanded consumption of animal tissues, and perhaps also increased hominin brain size (Féblot-Augustins, 1993; Ruff and Walker, 1993; Aiello and Wheeler, 1995; Potts, 1998; Braun and Harris, 2003; Sampson, 2006; de la Torre et al., 2014). Despite the precocious nature of LCT technologies when viewed against preceding African ESA industries, in the few Acheulean contexts potentially associated with hominin remains, LCTs have not been associated with the activities of anatomically modern hominins (Clark et al., 2003; Klein et al., 2007; Potts et al., 2018). Rather, in situ early Homo sapiens remains have been found exclusively in association with Middle Stone Age (MSA) assemblages.

Sites in the later Middle Pleistocene period—more recently referred to as the later ‘Chibanian’—∼400–200 ka (the later MP) are scarce in the southern and western Cape regions of South Africa relative to succeeding Stone Age occupations (Klein et al., 1999; Feathers, 2002; Herries, 2011, 2011; Braun et al., 2013; Caruana and Herries, 2021; Herries et al., 2022). The lack of archaeological evidence in this period is unfortunate as scientists widely assume that terminal Acheulean hominins made behavioral advances that forecast the later transition to behavioral modernity (Marean and Assefa, 2005). In fact, southern African occupations in MIS 9–6 are rare and patchily distributed regardless of industrial affiliation (Will et al., 2019), with understanding of hominin behavior in this period substantially reliant on the central interior localities of Kathu Pan and Wonderwerk cave (Fig. 1). Importantly, biostratigraphic estimates for the site of Duinefontein 2, located ∼180 km west of Montagu Cave (MC) on the western Cape coast, are inferred to be ∼300–250 ka, and the associated lithic assemblage is considered by some specialists to be potentially transitional between the ESA and MSA (Cruz-Uribe et al., 2003; Kuman et al., 2020). The scarcity of LCTs at Duinfontein 2 is, however, the very basis for the possible transitional attribution and, while the general presence of LCTs at this site is well established, the assemblage has never been published in detail. Additionally, classic Acheulean LCT assemblages have been recovered from well-contextualized layers at the site of Amanzi Springs dating—in different layers and locations—to ∼534–390 ka (Caruana et al., 2023). Yet, despite a small number of recently well-excavated localities, the archaeological record immediately preceding the emergence of the southern African MSA remains poorly resolved in terms of geochronology and regional variability in hominin behavior (Chazan et al., 2008; Porat et al., 2010; Key et al., 2021; Chazan, 2022), in some respects echoing the temporal and spatial uncertainties of the African hominin record associated with this period (McBrearty and Brooks, 2000; Galway-Witham et al., 2019; Roksandic et al., 2021).

Archaeologists have frequently associated later MP occupations of the interior regions of southern Africa—and occasionally also of the eastern African rift system—with the so-called Fauresmith industry (Tryon and McBrearty, 2002, 2006; Kuman et al., 2020). This term has been used variably to label assemblages in the terminal Acheulean, the earliest MSA, but more frequently to describe a mosaic industry with both ESA and MSA elements containing small LCTs (<12 cm in average length) that are relatively less numerous at the assemblage level than in the earlier Acheulean, large blades, rare prepared core flake production, and marginally trimmed convergent points (Goodwin and Van Riet Lowe, 1929; Mitchell, 2002; Porat et al., 2010; Kuman et al., 2020). What descriptions of transitional industries such as the Fauresmith have in common is the expectation of an archaeologically visible, gradual assembling of characteristic MSA technologies through the African later Middle Pleistocene, such as blades and points, and a concurrent loss of elements assumed to be relatively more reflective of archaic hominin behavior, such as LCTs. Within this trajectory of gradual innovative gain and concurrent loss of apparently maladaptive technologies, one might expect transitional industries with mosaics of Acheulean and MSA elements such as the Fauresmith (or the equivalents in equatorial African settings).

Interestingly, descriptions of the composition of the Fauresmith—as is also the case with several other southern African Stone Age entities—have tended to vary between researchers (Underhill, 2011; Wilkins, 2020) and the regional extent, geological determinants and chronology of the industry are still ambiguous (Humphreys, 1970). Similar challenges of compositional and geochronological consistency are shared by other supposedly ESA to MSA transitional industries such as the Sangoan and Lupemban. Recently, however, scientists have considered the label ‘Fauresmith’ important in describing assemblages that document the gradual technological refinement and overall decline of Acheulean technologies in the later MP of southern Africa, in favor of lighter and more diverse MSA toolkits (Wilkins and Chazan, 2012; Chazan, 2015; Kuman et al., 2020). It is widely assumed that these lighter MSA toolkits afforded hominins—argued by most scientists to be H. sapiens—key selective benefits over their Acheulean predecessors (Potts et al., 2018) and that, in the southwestern Cape, the ecological context of the Cape Floral Region (CFR) was one driver in the evolution and proliferation of later MSA technological behaviors (Marean, 2010). It is now clear that these lighter MSA toolkits were established to varying degrees in several regions of Africa by 500–300 ka (Wilkins and Chazan, 2012; Wilkins et al., 2012; Wilkins, 2013; Richter et al., 2017; Brooks et al., 2018); however, little is known about the tempo and process of behavioral evolution in contexts just preceding the onset of the MSA outside of eastern Africa. One reason for this gap in the southern African record is that the temporal and spatial extents of the Acheulean, including how these parameters vary relative to the local MSA, remain poorly documented.

Southern Africa has a relatively large number of MSA cave sequences dating to the last ∼150 kyr, yet integrating this record with the local Acheulean has been challenging due to a paucity of well-contextualized and continuous cave sequences (Deacon, 1998). This is unfortunate as the end of the Acheulean, and by implication the origins of the local MSA, may be key to exploring the emergence of behaviorally modern humans (Hublin et al., 2017; Richter et al., 2017; Potts et al., 2018). Montagu Cave, located in the mountainous interior between the south and west coasts of South Africa, is well positioned to contribute relevant data to this issue (Fig. 1). The site contains multiple, well-stratified, dense concentrations of artifacts associated with ochre and plant remains such as phytoliths as well as fruit and seed casings, reflecting some of the latest Acheulean occupations of South Africa's southwestern Cape. The CFR within which MC is located has been key in our understanding of the environments in which southern African modern humans evolved.

Here we describe renewed excavations at MC, focusing on stratigraphy and site formation processes, technological variability emphasizing LCT production, geochronology of the lower layers, and ecological context of the Acheulean occupations. The recovery of in situ assemblages from MC documents Acheulean Middle Pleistocene hominin occupation in the CFR during MIS 7, reflecting the adaptations of hominins in the period directly preceding the first documented appearance and proliferation of modern humans in the southwestern Cape.

Montagu Cave is a large cave formed within intensely folded sandstone formations of the Table Mountain Group, located approximately 160 km northeast of Cape Town, South Africa, on the flank of a southwest-oriented kloof (a steep-sided, vegetated, small valley) in the Langeberg mountain range of the Cape Fold Belt (Fig. 2a). The site is located close to the boundary of the modern year-round and winter rainfall zones (Fig. 1b) and, in accordance with aggregate regional paleo-vegetation archives, potentially experienced similar precipitation patterns for much of the last 300 kyr (Hijmans et al., 2005; Chase and Meadows, 2007; Dupont et al., 2021; Braun et al., 2023). Extremely fine-grained, primary sources of quartzite as well as perennial water and associated flora and fauna are available within 300 m of the cave today (Fig. 2a), suggesting minimal risk in the accessing of shelter, food, fuel, water, and high-quality stone, and that these resources may also have been available to hominins at a single location on the landscape for long periods of time in the past.

As previously mentioned, MC is located in the CFR, which has been relatively stable since the Late Miocene (Linder, 2003). This region has exceptionally high levels of biodiversity, with rich flora consisting of many endemics, and also contains a rich and diverse archaeological record pertaining to the last climate cycle (last ∼150 kyr; Dupont et al., 2011). The history of hominin adaptation in the CFR continues to yield important insights onto our understanding of hominin behavioral variability and evolution (Marean, 2010; Braun et al., 2021) and has been crucial in documenting some key contexts in which southern African modern humans evolved (Parkington, 2001, 2010; Texier et al., 2010). Archaeological and natural terrestrial archives implicating hominin paleoecology in this biome, however, are scarce prior to the Last Interglacial (Rossouw et al., 2009; Dupont et al., 2021; Braun et al., 2023). Montagu Cave is the only documented site in the West Montane part of the year-round rainfall zone. This part of the year-round rainfall zone is the richest subregion of the CFR (Cowling and Lombard, 2002) and is nested among a mosaic of Fynbos and Mesic Renosterveld vegetation that contains abundant endemic geophytes that could have been viable food sources for hominins (Vlok et al., 2005; Mucina and Rutherford, 2006).

Oxygen and carbon isotope records drawn from a single speleothem located ∼18 km west of MC documented climatic/biome stability across portions of the Middle Pleistocene period that MC was likely occupied, which was interpreted to underpin low rates of ecological change and correspondingly high present-day floral diversity in the immediate region (Braun et al., 2023). Broader inferences from Braun et al. (2023) would benefit from an increased sample of local paleoclimate records. That said, the implications of a relatively lower risk ecological context wherein resource availability may have been predictable over long periods of time suggests that MC may be well positioned to contribute to our understanding of the role of environmental variance in hominin adaptation (Potts, 1994, 1998), why specific behavioral strategies emerged or were retained in response to the effects of stable ecological parameters (Oswalt, 1976; Torrence, 1983; Collard et al., 2011), and the emergent stone artifact technologies resulting from this interplay (Braun et al., 2021; Iovita et al., 2021). Considering the proximity of MC to several well-dated MSA sites in the winter rainfall and year-round rainfall zones, aspects of the geochronology from MC in correspondence with contextual records of climatic variability may also contribute to our understanding of the pattern and pace of Middle Pleistocene behavioral evolution in the southwestern Cape (McBrearty and Brooks, 2000; Marean and Assefa, 2005).

The history of research at MC was initiated with the 1919 excavations by Houghten and Barnard, who were from the institution known then as the Cape Town Museum (Goodwin, 1929; Goodwin and Van Riet Lowe, 1929). More than 75% of the original archaeological deposits were apparently removed in this first excavation with unsystematic techniques (Keller, 1973), though reconstruction of the lateral extent of the 1919 excavation suggests this percentage is likely an overestimate (Fig. 2b). Keller's 1964–1965 excavations that followed were far more meticulous and, through manual bivariate piece-plotting (measuring X and Y), aimed to document stratified lateral distributions of stone artifacts in a portion of the remaining deposits, inferred then to be hominin living surfaces/floors (Keller, 1966, 1973). Although Keller's excavation focused on a smaller area of deposit in the cave than the 1919 campaign, his group recovered dense stone artifact assemblages from stratified contexts (n = ∼275,000 excavated lithics, although components were transported to foreign institutions and never returned to South Africa), and also left a portion of unexcavated deposit pertaining largely to the Acheulean sequence. Under the Ph.D. supervision of Prof. J.D. Clark, Keller's interest in MC was aligned with the spatial focus of African Acheulean research around that time (Clark, 1954, 1964; Isaac and Isaac, 1977) and aimed to document the organizational use of the shelter by hominins through exploring lateral artifact densities, focusing less on the taphonomic contexts of the assemblages recovered and their technological implications.

The sheer size of the Acheulean sequence at MC, both in terms of artifact density and original sediment volume, is a significant feature of the site, and is certainly unusual if not unique among known Acheulean sites in southern Africa. In short, two lengthy Acheulean sequences, with much the same technologies represented in multiple layers, are separated from one another by an archaeological hiatus. Above the second Acheulean sequence is an unconformity represented by roof collapse. The MSA and Later Stone Age (LSA) sequences follow this unconformity. The following description of the excavated sequence at MC draws on published terminology (Keller, 1966) relative to our on-site observations of remaining sections. Notably, here, the term ‘Layer’ refers to a sediment body with multiple internal stratigraphic entities that vary in color, compaction, and composition (essentially stratigraphic aggregates corresponding to a particular archaeological package of assemblages). Keller's excavations documented a sequence initiating with a lower Acheulean unit labeled by him as Layer 5, which started accumulating on the original bedrock of the cave. Layer 5 comprised multiple alternating black, white, and red organic-rich sand, silt, and clay lenses that dip toward the southeast (the cave mouth) and thicken in a westerly direction from the dripline toward the back of the cave, to a thickness of ∼2.5 m. The Layer 5 deposits closer to the mouth (the east) have been postdepositionally deformed due to diagenesis and consequent loss of volume. Layer 4 overlies Layer 5 and is comprised of a thick homogeneous and archaeologically sterile ferruginous sand (>1 m thick in certain parts of the site). Following this sterile unit (Layer 4) was another shorter Acheulean archaeological sequence comprising black, white, and red laminated bands with sedimentological composition, compaction, and color variability similar to those of the stratigraphy of Layer 5, indicating a similar formational process. This upper Acheulean sequence was labeled by Keller as Layer 3 and, like Layer 5, dips towards the overlying boulders at the cave mouth. Both Acheulean units (Layers 5 and 3) contained dense, largely flat-bedded, classic Acheulean artifact assemblages (>550 LCTs were recovered from in situ stratigraphic contexts), which correlated with our observations and spatial data documenting these layers recorded in 2014–2017 (Fig. 3). Both Acheulean units, according to Keller, also had structured variability in the vertical density of accumulated anthropogenic materials, meaning concentrated lateral extents of artifacts interpreted as living floors, interspersed with relatively decreased more randomly distributed finds.

Directly overlying the upper Acheulean was Layer 2, which broadly contained MSA assemblages that extended in the vicinity of the dripline and sediment trap at the mouth of the shelter. No sedimentary hiatus was documented by the previous excavators between Layers 3 and 2; however, our work at the site documented a significant roof-fall event extended toward the dripline, between the accumulation of the latest Acheulean and youngest MSA in the cave (Fig. 2c). Due largely to substantial numbers of backed artifacts (so-called segments and trapezes) throughout the MSA levels, Keller attributed all of the MSA to the ‘Howiesons Poort’ industry (Volman, 1981); however, he also suggested in his dissertation that there were some affinities of the lower MSA at MC with the now obsolete quartzite-rich ‘Mossel Bay’ industry (Keller, 1966), which shares a small number of technological features with earlier MIS 5 assemblages from the sites of Elands Bay Cave, Diepkloof Rockshelter, and Klasies River Mouth (Porraz et al., 2013). Layer 1 was the relatively thin uppermost unit documented by Keller and contained LSA lithics. The LSA deposits were closer to the dripline and today are completely removed from the site. Surprisingly, during the removal of backfill from previous excavations (which contained substantial quantities of ex situ lithics) and during our own excavation, our team did not identify any unequivocal LSA artifacts. The lower LSA at MC was attributed by Keller to the ‘Robberg’ industry due to the frequency of bladelets, micolithic scrapers, and the prevalence of characteristic bullet-shaped bladelet cores (Porraz et al., 2016; Low and Mackay, 2018; Arthur, 2022), and the upper LSA to the Wilton (Keller, 1973; Mitchell, 2002). The raw material spectrums in Keller's named archaeological units were all dominated by local high-quality quartzites; however, the frequencies of silcretes and cherts were relatively higher in the MSA and LSA. Carbon 14 charcoal samples from the MSA levels, processed soon after the 1964–1965 excavations, yielded ages ranging from ∼23.2 to 45.9 ka (Keller, 1973), which today cannot be considered reliable, given recent well-established yet contested luminescence chronologies published for the later MSA of the southwestern Cape (Jacobs et al., 2008; Tribolo et al., 2013). The Acheulean stratigraphic layers at MC were not previously dated. Our excavations documented abundant remaining in situ deposits of Keller's Layers 3–5, as well as pockets of ephemeral MSA deposits in less optimal taphonomic contexts, in amongst the boulders comprising the sediment trap.

Apart from recent fieldwork at Amanzi Springs in the Albany Thicket Biome, and research at older, largely deflated, open air sites ∼180 km to the west of MC (Cruz-Uribe et al., 2003; Dietl et al., 2005; Klein et al., 2007; Archer and Braun, 2010; Kandel and Conard, 2012, 2012; Herries et al., 2022; Caruana et al., 2023), there are few contexts to explore how Acheulean hominins survived in the unique environments of the southwestern cape, and how their adaptations contrast with the much larger corpus of data on behaviorally modern hominins associated with the MSA of this region. It has been widely argued that MSA technologies offered later Middle Pleistocene hominins selective benefits relative to Acheulean technologies, particularly in the acquisition of resources in rapidly changing and unpredictable eastern African environments (Brooks et al., 2018; Potts et al., 2018). The inverse prediction is that certain Acheulean technologies may have actually enhanced extractive foraging abilities and/or optimized energy expenditure in low-risk contexts where resource availability was predictable for substantial periods of time, and group mobility was correspondingly low. This possibility perhaps warrants consideration in contexts where classic Acheulean technologies were practiced by hominins unexpectedly late, such as in the southwestern Cape, and where long-term Middle Pleistocene climatic/biome stability appears to be reasonably well established (Dupont et al., 2021; Braun et al., 2023).

A renewed field campaign at MC between 2014 and 2017 sought to recover a relatively small archaeological sample with modern excavation and digital recording techniques, to better contextualize the older systematically excavated collections and to generate detailed data on hominin occupation of the southwestern Cape during the later Middle Pleistocene. Prior to commencement of fieldwork, using a combination of historical photographs and modern three-dimensional (3D) reconstructions of the remaining deposits in the cave, we established that substantial deposits from the Acheulean horizons were still present near the mouth of the cave, sealed below a boulder collapse (falling of the roof overlying the cave mouth) that occurred between the Acheulean and MSA occupations of the site.

Here, we present the first results from the excavation, geoarchaeology, site formation, paleo-vegetation, as well as geochronology of these lower Acheulean deposits, focusing on the levels we were able to date: 21 and 22, which are identified here using Keller's (1966) IDs as the luminescence samples derive from the freshly exposed 1964 sections, which we subsequently excavated in several locations. The excavated levels and time period represented at MC proved challenging to date with available luminescence techniques. The current paper therefore focuses on occupations in the two stratigraphic levels from the Acheulean sequence, which were dateable with luminescence techniques, to explore links between site formation and behavioral data. As the dated levels occur toward the base of the stratigraphic sequence (Fig. 4) and site formation suggests a probable rapid accumulation of Acheulean deposits, it may be reasonable to propose that the acquired ages provide a reflective age estimate for the Acheulean occupation of MC in general.