Riddles wrapped inside an enigma. Lupemban MSA technology as a rainforest adaptation: revisiting the lanceolate point

Abstract

The Central African Stone Age is very poorly known when compared to the higher-resolution records of East and Southern Africa. Early Stone Age (ESA) archaeology is effectively absent from the rainforest zone, with the early Middle Stone Age (MSA) Lupemban industry representing the earliest sustained archaeological signature. Uranium-series dates of approximately 265 ka BP for the Lupemban at Twin Rivers (Zambia), although queried, suggest a precocious late Middle Pleistocene dispersal of early Homo sapiens into the equatorial rainforest belt. Lupemban palaeohabitat interactions and attendant behavioural and technological repertoires are key to its evolutionary significance, but investigation is hampered by the widespread vertical disturbance of stratigraphic profiles and the formation of ‘stone-lines'. The Lupemban takes in a range of implement types and technologies, including core-axes, prepared core technology (PCT) points, blades and backed blades. But it is the elongated bifacial lanceolate point—some exquisitely made and many exceeding 30 cm in length—that defines the industry. Remarkably, unequivocal examples of these iconic artefacts have never been the focus of detailed techno-typological scrutiny. In this paper, I advance understanding of the Lupemban by initiating a re-consideration of lanceolate points at Kalambo Falls, Zambia, and discuss their implications for the Lupemban's evolutionary significance.

This article is part of the theme issue ‘Tropical forests in the deep human past’.

1. A precocious African origin of rainforest foraging?

Historically, archaeological research into the antiquity of rainforest foraging in Africa reflects a series of paradigmatic swings. As background, the inhabitation of the African forest belt has long been recognized as the ancestral condition of the primate lineage, and to be a critical feature of early hominin evolution [1]. In the mid-twentieth Century it became widely accepted that aridification caused the fragmentation of forests and the expansion of savannah ecosystems across East and Southern Africa during the early Pleistocene, with open grasslands constituting the essential impetus and environmental backdrop for most of human biological and technological evolution [2]. The hominin settlement of rainforests was thought to have only been re-established by Homo sapiens in the recent Holocene. Put simply, jungles have most frequently been considered an evolutionary sidenote, offering mere punctuation to the human story.

The eminent archaeologist J. Desmond Clark repeatedly contended from the 1950s onwards that mobile human foragers had lived long-term within dense, closed canopy rainforests at least as far back as the African Middle Stone Age (MSA) [3–9]. The basis for his contention was an apparent correlation between the contemporary Central African rainforest belt and archaeological sites featuring stone tools of the early MSA Lupemban industry [10]. Based on emergent radiocarbon (¹⁴C) dates, the transition from the handheld Acheulean tools of the Early Stone Age (ESA), to hafted MSA technologies—including the Lupemban—across Sub-Saharan Africa was considered to be rooted in the Late Pleistocene, and implied an approximately 40 kya initial origin of rainforest foraging [11]. This was the dominant archaeological paradigm throughout the 1960s and 1970s—the ‘golden era’ of Central African Stone Age research—but was increasingly queried in the 1980s by a series of refinements in both archaeological and ethnographic data.

The chrono-stratigraphic integrity of Central African Stone Age sequences was cast into doubt in the late 1970s by Cahen's discovery of the extreme (greater than 1 m) vertical dispersal of refitting lithic artefacts in separate horizons at Gombe Point [12]. Subsequent recognition of the widespread effect of several mechanisms of bioturbation on weakly consolidated sediments, including the action of tree roots [13], termites [14] and repeated wet–dry cycles [15], means that the association of sub-surface artefacts with one another—as well as with datable materials (e.g. charcoal) and palynological samples—is frequently highly unreliable. Concurrently, ethnographic debates and data point to the significant challenges closed-canopy rainforest habitats pose for contemporary foraging communities. Rich in biomass but poor in protein and carbohydrates, it was suggested that African forests present impediments to long-term settlement that demand specialized knowledge and extractive technologies; and, even then, settlement may be possible only when diets are augmented with externally sourced foodstuffs [16,17].

The paradigmatic pendulum began to swing back in the 1990s when advances in radiometric dating shifted the timeframe of the ESA–MSA transition back into the past by an order of magnitude [14]. This included a series of Uranium-series dates for the Lupemban deposits at Twin Rivers in South Central Africa (Zambia) that revealed an age range of approximately 270–170 ka [18] for a sequence previously dated by ¹⁴C to over 33,200 ka [19]. There has been some debate over this new chronology [20,21], but it remains as the most reliable direct assessment of the timespan of the Lupemban. These dates provided the basis for Barham's [22] postulation that the general distribution of Lupemban sites might reflect the incursion by early Homo sapiens into the rainforest belt during MIS 7. Current evidence for the Lupemban as a specific rainforest adaptation remains equivocal [23–25], but a range of new archaeological and ethnographic data from further afield in Africa, and in Asia, lends the industry and this question renewed paradigmatic relevance.

Recently, new data on the so-called ‘Wild Yam Question’ have suggested that underground storage organs may enable independent foraging within African rainforests [26]. Concurrently, evidence for a human presence in the rainforests of Sri Lanka by at least ca 48 ka [27,28] seems to have settled the immediate question of the penetrability of rainforests by humans in a global context, but the precise picture in Africa is less clear. At Panga ya Saidi (Kenya), archaeological remains dated to approximately 78 ka offer the best clue but are set in an rainforest ecotone [29], thus indicating, as both Cornelissen [30] and Taylor [25] have suggested, a partial adaptation to closed canopy conditions. If they exist as part of an undisturbed sequence and their purported ∼800 ka age can be reliably demonstrated, then reports of potential artefacts at Elarmékora [31] may yet have paradigm-shaking implications for the deep antiquity of hominins in lowland Central Africa.

2. The Lupemban in Central African context

The early MSA Lupemban industry has been considered a circumstantially strong candidate for the initial settlement of the African rainforest belt since the 1950s [22]. Its overall distribution is said to be roughly coincident with the contemporary extent of the Congolese rainforest belt [8,25]. Without a single occurrence in primary (undisturbed) archaeological context, precise definition of its technological content has proven impossible, but based on more intact sequences on the margins of the region (e.g. Kalambo Falls, Twin Rivers, Muguruk), it is considered to be defined by the large bifacial lanceolate points that are the focus of this paper. Considered as fossiles directeurs, they are absent from the Sangoan industry that constitutes the first evidence of human activity at some sites in Central Africa but is extraordinarily poorly known and weakly defined [32]. Heavy-duty core-axes and picks in the Lupemban suggest its ancestral link with the Sangoan and the late Acheulean, in which they also feature [9]. More broadly, the Lupemban seems to bear witness to the initial and widespread appearance across Central Africa of prepared core technology (PCT), blades, and at some sites, geometric backed blades [33], while at Twin Rivers, it is also associated with large quantities of anthropogenically accumulated pigments [34].

A dearth of intact chrono-stratigraphic sequences sampling the Lupemban and later MSA across the Congo Basin and adjacent areas has severely impeded the derivation of site-based palaeoecological samples that can test for technological associations at the scale of human communities and individual agents. Samples of varying reliability and resolution come from just five sites. More reliable are the records from Twin Rivers (open woodland-bushland [35]) and Kalambo Falls (Miombo woodland [36]). At all sampled sites from lowland Central Africa—Mufo (savannah grassland/woodland [4]), Mosumu (rainforest [37]) and Ndjole (rainforest [38])—palynological samples are limited and relate to vertically disturbed ‘stone-line’ sequences that undermine any lithic-pollen associations. This limited dataset has led to more generalized attempts to determine the industry's palaeoenvironemental correlations, based on wiggle-matching the distribution of Lupemban sites with rainforests under expanded (interglacial) and contracted and fragmented (glacial) marine isotope stages (MIS) ([22,39]; and see [25], p. 280]. The problem with these approaches is that they necessarily assume site (pene)contemporaneity [22] and the attribution of all sites published as ‘Lupemban’ to the industry based on limited typological data.

The elongated symmetrical bifaces found in the Lupemban have historically been called lanceolate (or foliate) points by Anglophone archaeologists, and pointes foliacées by Francophone workers, but more loaded terms such as lancehead [40,41] and assagai [42] have occasionally been used. Where known from intact sequences they are rare, but in spite of this and their morpho-metric variability (figure 1), they have been treated as fossiles directeurs. Problems arise in their use as markers to identify, track and delimit the Lupemban because their definition has rarely exceeded the level of typological description, even from better-known and stratified sites. The complete absence of basal thinning or tanging of Lupemban lanceolate points is said to distinguish them from the North African Aterian [43,44], and they are superficially more similar to the bifacial Still Bay points of later MSA in South Africa. These points range in length from 37–76 mm at Hollow Rock Shelter [45] and 34–74.5 mm at Blombos Cave (South Africa), where they were axially hafted and retain tip damage from impact [46]. But Lupemban lanceolates are markedly different; they regularly exceed 250 mm in length and may reach 350 mm (figure 1), an order of magnitude larger than many Still Bay examples.

Figure 1.

Variability in the size and shape of large, elongate lanceolate points that are the fossile directeur of the Lupemban MSA across its purported geographical range. Artefacts relate to sites: (1) Mosumu, Equatorial Guinea, (2) Porte de l'Okanda, Gabon, (3) Gombe Point, Democratic Republic of the Congo (DRC), (4) Lupemba, DRC, (5) Musolexi, Angola, (6) Kalambo Falls, Zambia, (7) Masango, Burundi, (8) Nsongezi, Uganda, (9) Muguruk, Kenya and (10) Lodjo, DRC. After [37] but all artefacts shown to scale. (Online version in colour.)

Beyond these broad morpho-metric comparisons, the geographical and temporal delineation of the Lupemban remains extremely difficult, complicating efforts to understand its distribution, potential palaeoenvironmental associations and relationships with other MSA industries. The proportional rarity of lanceolates means they are absent in some assemblages that otherwise share technological affinities with the industry (e.g. Peperkorrel, Kamoa). Similarly, the presence of Lupemban-like lanceolates at younger sites (e.g. Birimi [47, pp. 79, 152]) could indicate the industry's late persistence as part of the young West African MSA [48], or mere technological convergence [49]. In many historical cases, textual reports of typological ‘lanceolates’ were considered sufficient evidence for Lupemban assemblage attribution, but this descriptor can include both very large bifacially flaked artefacts and diminutive unifacial or minimally flaked, foliate-shaped PCT points.

A lack of longer-ranged dates and extremely poor stratigraphic control at Central African MSA sites mean there is almost no understanding of the timespan of the Lupemban, or technological change between the early MSA and Later Stone Age (LSA). While Clark [4] identified lanceolates in the Sangoan of northern Angola, this is a semantic issue; expanded use of the term Lupemban has subsumed some technological elements previously considered Sangoan. Lanceolates do however appear to be a consistent feature of the MSA in northern Angola [4]. Here Clark identified broad changes in the Lupemban as it graded into a later MSA Lupembo–Tshitolian industry; foreshadowing the appearance of the sub-regional LSA Tshitolian industry. Lupembo–Tshitolian lanceolates were routinely made by pressure technique and exhibit serrations, basal thinning and tanging. Elsewhere, models of nuanced progressive change in Lupemban lanceolate morphology have been proposed [4,9,50], but these lack any firm stratigraphic foundation. At Muguruk (Western Kenya) lanceolates occur in a mixed Sangoan–Lupemban ('Ojolla') industry that is overlain by a horizon of non-diagnostic MSA material [51], indicating that here they were not a consistent diachronic feature throughout the MSA. In South Central Africa, they are present in the Twin Rivers sequence dated to approximately 270–170 ka [20,21], but are absent from the approximately 120 ka primary MSA deposits at the nearby site of Mumbwa Caves, suggesting that here foliates probably do not extend into the Late Pleistocene [52].

As potential markers for the Lupemban and the first sustained settlement of the Central African rainforest belt, refined understanding of lanceolate points may eventually enable improved spatio-temporal tracking of the industry and its distribution. Here I refocus attention on the Lupemban lanceolate point as known from the most securely stratified assemblage from Kalambo Falls, Zambia. Though not without limitations [53], as I have previously shown this sequence has to-date provided the only Lupemban assemblage in isolated stratigraphic context [25], and it thus provides the only unequivocal Lupemban assemblage currently known. It is the logical starting point for any re-assessment of the industry, against which future inter-site correlations might be drawn. The work presented here lays the groundwork for future studies examining its distribution and spatio-temporal variability more precisely.

3. Lupemban lanceolate points at Kalambo Falls (Zambia)

Of the more than 25 000-piece Lupemban lithic assemblage recovered from Kalambo Falls between 1953 and 1966 [9,41], detailed information is available for only 10 of the 14 lanceolates reported ([9], table 1). Of these, just four artefacts could be re-located for this study, and only three were available for detailed analysis. Their technological analysis (figure 2) is based on the well-established principle of scar direction and superposition (i.e. truncation/cross-cutting) [54]. Post-depositional alterations mean sequencing is only possible between broad categories: initial shaping scars (wide, shallow and highly invasive soft hammer scars: often traverse the midline and always superimposed by subsequent scars); secondary shaping scars (elongated, usually laterally truncated scars that may feather, step or hinge terminate [55]) and final trimming/resharpening scars (non-intrusive hinge or step terminating scars that re-position and/or regularize the perimeter).

Table 1.

Details of lanceolate sample from Kalambo Falls, Zambia. Specimens included in this study are highlighted.

^aBased on codes written on artefact surfaces.

^bIt is impossible to distinguish between Rubble horizon sub-units (Ic (i), Ic (ii) and Ic (iii)) based on the stratigraphic references written on artefacts.

^cBased on post-excavation scars, where possible.

Figure 2.

Technological analysis of key Lupemban lanceolates from Kalambo Falls, Zambia. For artefacts (a) and (b) dorsal faces are on the left side, for artefact (c) the left side is considered Face 1, right side Face 2. Unshaded scars on (b) cannot be confidently ascribed to any of the three scarring phases. (Online version in colour.)

(a) . Manufacture

Lupemban lanceolates are described as the result of bifacial façonnage reduction using invasive soft-hammer retouch, and as being either single- or double-ended. Typically, their entire lateral edges are carefully retouched and highly regularized. Among other characteristics, this is said to differentiate them from Lupemban core-axes and picks, which retain significant thickness and weight in their mesio-proximal sections, have irregular and steep lateral edges and distal margins formed either to a point or a transverse working bit. Tools called 'lanceolates' across Central Africa, however, incorporate considerable morphological variation (figure 1), and in individual cases, the distinction of core-axes and lanceolates may not be agreed upon [56, p. 178]. Clark [5] noted that the manufacture of some later examples from Angola involved retouch by punch technique, and that ‘the bifaced foliate points not infrequently show serrated edges' [4, p. 157].

Both Kalambo Falls lanceolates made of feldspathic quartzite retain complete, non-cortical striking platforms at their proximal ends. This shows their reduction from large end-struck flakes (figure 2a,b), themselves removed from the medium–large cobbles of this material found in the local basin [9]. The presence of a fault with cortex on the dorsal aspect of A1/56/III/e8 (figure 2a) indicates it derives from an early stage flake. Ventral flaking is shallow and flat, while dorsal scarring is steeper, producing somewhat asymmetric lenticular sections. The origin of the silicified mudstone specimen A3/56/I c2/3 (figure 2c) is less clear; it shows longitudinal curvature indicative of a flake blank, but invasive bifacial retouch of its proximal end has removed evidence of any platform. Production of large thin bifaces from prismatic blocks of Kalambo Falls mudstone is extremely difficult due to their limited width (approx. 10–15 cm) and multiple internal (transverse) faults. More likely this specimen was shaped from a nodule in the manner of mudstone core-axes. One additional mudstone lanceolate (see Fig. 4. 20 : 6 in [9]) may reflect core reduction but is unfinished and perhaps better considered as a core-axe.

In all three cases, initial shaping consists of very invasive soft hammer scars struck from opposing lateral edges towards the midline, which they sometimes traverse. This strategy, and the absence of longitudinally struck shaping scars, enabled the imposition of bifacial symmetry while maximizing artefact length. Edges adjacent to tool proximals show reshaping by superposing secondary retouch, involving shorter and more elongated removals that follow existing facial arêtes. There is no evidence of attempts to impose specific hafting modifications, for example basal thinning. The occurrence of secondary shaping scars in run-together patches on lateral edges appears to have been key to their regularization; a more limited presence on Kalambo Falls core-axes and picks seems to generate their greater sinuosity [9]. The importance of secondary shaping is apparent on figure 2a, where the greater majority of its edges have been formed using this approach. In figure 2c, scarring is patterned as an asymmetric unifacial reduction in which a lenticular or rhomboidal cross-section is maintained by shaping from opposing margins onto either face. Trimming or final retouch series scars are typically non-invasive and occur in superposed patches, often run together. On A1/56/III/e8 and A3/56/I c2/3, scar frequency—particularly involving more fine trimming retouch—is notably greater at and symmetrically adjacent to the tip, indicating that these lanceolate's sharp lateral edges provided structural support for these more important parts of their perimeter (see below) [9]. All three artefacts show a clear pattern of progressive lateral retouch, with A3/56/I c2/3 in particular showing problematic step fracturing. This seems to reflect the fact that the relative proportions of initially wider and thinner forms (e.g. figure 2a) became increasingly difficult to maintain with sequential retouch, their lenticular cross-sections becoming thicker and less easy to maintain, even as elongation was retained. Apparent reasons for tool discard vary; the transverse break on B2/II/L3 would necessitate significant artefact reworking (shortening), while the already thick profile and limited remaining width of A3/56/I c2/3 provided little scope for further resharpening and use. The reason for the discard of A1/56/III/e8 is less clear, since it provides ample latitude for reshaping and resharpening.

(b) . Function

Hypotheses of Lupemban lanceolate function have focused either explicitly [3, pp. 156–157] or implicitly [6,50,57] on their potential use as the business end of composite spears that were either thrown or thrust to kill large terrestrial game. Both Turnbull [58] and Janmart [59] have reported the use of morphologically similar iron tools for hunting wild boar and forest elephant by the Mbuti:

‘In the Ituri Forest, the Pygmies use short lances with a very large and broad iron blade (ca. 30 × 20 cm) mounted on a short (ca. 75 cm) and thick handle made of hardwood…The hunter sneaks underneath the standing elephant and thrusts the spear upward into its soft belly with a lightning-quick movement’. [59, p. 147, original emphasis retained.]

The use of extremely large and elongated stone-tipped weapons would intuitively appear equally appropriate for this task. Although not without shortcomings [60], Wilkins et al.'s [61] interpretation of tip fractures on Fauresmith points from Kathu Pan 1 as impact damage resulting from their use as thrust composite hunting weapons might be taken as circumstantial support for a widespread origin of cooperative hunting using thrusting armatures in the early MSA, including in the Lupemban. However, while the use of lithics as armatures implies obligate hafting, it is important to de-couple the idea of composite lanceolates as synonymous with hunting activities; there is space to consider whether their socketing into handles of wood or bone might have conferred an advantage as cutting implements, or even as a kind of machete. Handheld digging and cutting activities are mentioned by McBrearty [51, p. 413] as alternative hypotheses for the function of Lupemban foliates, but it is equally plausible they were hafted for such purposes.

In spite of their similar shape, when compared to other MSA hafted bifacial points [45,62], the size of Lupemban lanceolates suggests a radically different functional envelope. It has been argued that morpho-metrical attributes, particularly tip cross sectional area (TCSA) and tip cross sectional perimeter (TCSP) [63,64], can discriminate projectiles and weapon delivery systems through comparison with ethnographic tools of known function. Calculated TCSA values for the Kalambo Falls Lupemban lanceolates presented here are based on Sisk and Shea's rhomboid measure for bifaces, and fall respectively at 718 mm² (figure 2a), 520 mm² (figure 2b) and 350.2 mm² (figure 2c). Their TCSAs drastically exceed the ranges that Shea [63] provides for ethnographic arrowheads, dart tips, spear tips, and even experimental thrusting spear tips (79–257 mm²). They are also greater than the TCSA values for MSA points from southern Africa collated by Lombard [65: Table 5], the greatest of which are those from Kleinhoek 1, which reach a maximum of 322 mm² but are still markedly lower than smaller Lupemban lanceolates. The unifacial MSA points from Kathu Pan 1 (South Africa) are of particular interest, given their ca 500 ka age and tip fractures interpreted as thrusting impact damage [61], and these have maximum TCSAs of 299 mm². Of broader technological similarity are the bifacial Still Bay points from Umhlatuzana Rock Shelter [66], the largest of which is described as a ‘greater than 200 mm² outlier’.

It is worth reiterating that the metrics of the Kalambo Falls examples shown in figure 2 are entirely unexceptional for this Lupemban artefact class, both at this site [9] and more broadly (cf. figures 1, 2 and 3). Although still to be calculated for other incontrovertibly Lupemban assemblages, a range of approximately 750–350 mm² appears representative. It should be acknowledged that as well as gross size, morphology and TCSA, weight is also a critical determinant of projectile suitability, with heavier tips requiring more substantial handles to counterbalance and maintain aerodynamic stability [67]. A direct relationship between ethnographically calculated TCSAs and prehistoric projectile use is queried by Australian examples [68], showing that calculated ranges do not always correlate with the true aerodynamic limits or the optimization of weapons systems [69]. However the attributes of these representative Kalambo Falls examples fall so far outside those expected of serviceable projectile armatures that this use would appear very unlikely. A function as composite thrusting spears is more plausible.

Figure 3.

Examples of tip damage on Central African lanceolate points. The presence of mesio-distal transverse snaps and bending step-terminating scars on distal portions provides an initial indication of impact damage, but do not prove it. The lateral break on the left-most piece occurred after excavation and provides a cautionary note of the over-interpretation of lateral snaps. (1) Gombe Point, DRC (image permission courtesy of the Royal Museum for Central Africa, Tervuren); (2) and (3) Lukala, DRC; (4) Kanda Kumbi, DRC; (5) Kalambo Falls, Zambia; (6) Furi I, northern Angola. (Online version in colour.)

The reduction of Kalambo Falls lanceolates speaks to the issue of their function, albeit obliquely. The change in planform on the right lateral of specimen A1/56/III/e8 (figure 2a) has resulted from a pattern of unifacial semi-invasive retouch scarring not applied elsewhere on the piece. Their full negatives reflect a final stage of production intended to resharpen and reshape the tools' distal portion, and it is possible that the abrupt lateral cessation of this retouch reflects an approximate haft limit, from the artefact's modification in a haft (see fig. 9 in [46]). Less clear is the precise intention of the resharpening: whether to reform the distal tip and adjacent lateral edges to improve tip penetrability or to rework the lateral edge for its use, each provokes a different techno-functional inference as to the intended working edge.

The critical transverse break on artefact B2/II/L3 was probably the reason for its abandonment. Lithic use–wear research defines transverse snaps with spin-off fractures as one type of diagnostic impact fracture (DIF) [70], and though non-diagnostic, it is clear that lateral breaks are frequently produced during impact motions [62,71,72]. Snaps can also result from endshock during manufacture, but this is unlikely based on the author's experience with knapping Kalambo Falls quartzites. The question then remains: is this break functionally related, or not? More broadly, there is notable—if circumstantial—evidence of numerous Central African lanceolates that feature transverse tip breaks (figure 1: 6, 8, 9; figure 3: 3–6), bending step-terminating fractures originating at the tip (figure 3: 2, 6), and apparent bifacial spin-off fractures (cf. fig. 1: 5 and fig. 1: d2 in [73]). Dedicated use–wear and experimental research is needed to systematically examine these possibilities further. Until then, it remains possible that these specimens suffered accidental damage (due to manufacture, dropping, etc). The distal portions of lanceolates are their weakest points and more susceptible to fracture, however, the damage in many cases necessitates a highly idiosyncratic contact and other, non-diagnostic tip damage patterns have to date not been observed.

4. Discussion and conclusion

Examples of unequivocally Lupemban lanceolates are here shown to be dynamic artefacts. Similar to other MSA artefact classes (e.g. core-axes), their manufacture did not follow a single trajectory and neither is it likely their function was monolithic [74]. Based on the detailed analysis of a very small sample from the best-stratified Lupemban sequence, in the early MSA these artefacts were produced using soft hammers from large flake blanks, but also possibly from thicker blocks. A repeated pattern of opposed platform shaping and thinning is evident, with abandonment occurring due to presumed morphologically induced functional inefficiencies—either breakage or a thick cross-section/steeper profile. The absence of any pressure flaking/punch technique or serration on this sample or elsewhere in the MSA at Kalambo Falls [9] shows this technique is not a distinguishing feature of the Lupemban, at least in South Central Africa. It could yet prove a sub-regional variant, but based on Clark's work in northern Angola, it more likely emerged as a Late Pleistocene adjunct [4]. Future work on lanceolate techno-morphology would benefit from comparison with other better-stratified sites (e.g. Muguruk) to establish potential design rule-sets [75].

Initial TCSA calculations show larger lanceolates (especially those greater than 15 cm) are very unlikely to have functioned as hafted projectile weapons. While their cross sections also fall outside the expected range for hafted thrusting armatures based on experimental and historic correlates, this is a hypothesis still worthy of use–wear experimentation: the maximum TCSA ceiling for such thrusting weapons is not defined [65]. As a proxy for projectile weapon suitability, TCSA values cannot replace detailed use–wear analysis, and the presence of tip damage on several examples indicates possible impact use. Moreover, complex technological innovations, like biological structures, do not evolve instantaneously in an optimized form—Lupemban lanceolates may have emerged as multi-purpose bifaces but were co-opted as serviceable hunting tools. If, as hunting weapons, they conferred even a slight but important foraging advantage, they may latterly have become more refined as an idiosyncratic regional response to cultural, environmental and functional opportunities. That in Central Africa hunting weaponry originated early and gradually through the hafting of lanceolate bifaces, rather than PCT points, is also worth considering. Confirming their potential utility as medium-range hafted stabbing weapons, including within rainforest, woodland or mosaic forest-savannah biomes will require the combination of high-resolution functional (use–wear and/or residue analyses) and site-based palaeoenvironmental datasets relating to well-stratified samples that unequivocally relate to the Lupemban MSA.

Scaling up inferences drawn from the very small sample analysed here is problematic, but unavoidable. The focus on Kalambo Falls as the best-stratified [25], certainly very large and arguably best-reported Lupemban assemblage nevertheless resulted in the relocation of just three excavated lanceolates. This reflects a wider problem: at present, knowledge of the Lupemban's composition, distribution, chronology and habitat preferences varies at different observational scales, indeed in a way analogous to Heisenberg's [76] famous uncertainty principle. Analytically, details of its composition, chronology and paleoenvironment (i.e. ‘position’) and its distribution (i.e. ‘momentum’) cannot all be known simultaneously. Bottom-up approaches focusing exclusively on high-resolution data from sites and assemblages with better chrono-stratigraphic, techno-typological or palaeoenvironmental resolution result in small samples and force a collapse in any meaningful regional comparisons [25]. Conversely, top-down views of the industry can reveal its generalized distribution and overall makeup, but only by forgoing any real precision in its age, content and palaeoenvironmental associations.

The techno-morphology and proposed function/s of lanceolate points do not link them to any particular palaeoenvironmental profile. Indeed, whether hafted or not, they may just as well have been used in open savannahs or multifarious habitats, as within rainforest and woodland habitats. Putting caution aside, however, if the Lupemban as a whole emerged on the margins of Central Africa approximately 270 kya and continued until 170 kya as indicated from Twin Rivers [21,22], then these tools may yet prove to have offered a solution to big game hunting that was developed by early Homo sapiens to exploit MIS 7 rainforests, as part of a wider multi-functional Lupemban toolkit [22]. Just one alternative is that they instead provided a means by which to forage among adjacent woodland, savannah and rainforest ecotones on the margins of Central Africa, with their apparent diachronic refinement reflecting a long process of repeated, variable scale incursions into closed canopy rainforests. As the expanded rainforests of MIS 7 gave way to the retracted conditions of MIS 6, Lupemban tool users may have pursued these ecotonal conditions and incidentally dispersed into lowland Central Africa over the course of hundreds or even thousands of human generations. Regional and sub-regional pulses of palaeoenvironmental expansion and contraction through the late Middle and Late Pleistocene would then provide the backdrop for an increasing familiarity with, and persistence within, closed-canopy rainforest environments. In the absence of refined chronological, palynological and techno-typological datasets, this view remains highly speculative but plausible as a broad and long-term model for the origins of rainforest foraging in Central Africa.

Data accessibility

This article has no additional data.

Authors' contributions

N.T.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, writing—original draft, writing—review and editing.

Competing interests

I declare I have no competing interests.

Funding

I received no funding for this study.