Late Middle Stone Age Behavior and Environments at Chaminade I (Karonga, Malawi)
The African Middle Stone Age (MSA, typical range ~ 320–30 ka) has been the subject of intense research interest in recent decades as a culture-chronological Unit associated with the emergence and dispersal of our species. Recent results of this work have shown that sites designated as “MSA” contain common approaches to lithic reduction, but that within this rubric, there is much diversity in overall assemblage characteristics and the timing of their appearance across the continent. As researchers recover more data from more sites, especially from undersampled geographic regions, this more complex picture of the MSA reveals technological and other behavioral diversity in early modern human populations that may inform about the ultimate success of our species. Here we add to this growing database by describing the environmental context and characteristics of two concentrations of stone artifacts from the late MSA (~ 43–21 ka) open-air locality of Chaminade-I (CHA-I), near the town of Karonga in northern Malawi. The CHA-I lithic artifacts show a flexible approach to stone tool production and use that is common to assemblages in Karonga but distinctive from MSA sites reported elsewhere. Radial and minimally reduced cores typify an unelaborated lithic assemblage, in which raw material choice is driven by toolstone clast size and shape rather than preferential use or treatment of specific materials, as found in MSA assemblages in the East African Rift, South Africa, and the North African coast. Lithic reduction at CHA-I took place within a woody, riparian context embedded within a more open woodland landscape. Most artifacts occurred in near-channel sandy deposits dated to ~ 41 ka, and were buried under alluvial fan deposits that began aggrading by at least ~ 21 ka and continued beyond ~ 5.5 ka within a grassy, open landscape. The site’s late MSA age and lack of elaboration in lithic technology challenges straightforward ideas of increasing complexity in human technological behavior over time and provides important insight into the diversity of MSA technologies and the environmental conditions in which they existed.
KeywordsSite formation East-Central Africa Stone tools Vegetation change
The Middle Stone Age (MSA) of Africa is variably defined along chronological and technological terms, as both a time period in which Homo sapiens emerged and proliferated in Africa and by a set of behavioral adaptations documented in approaches to flaked stone tool manufacture (Clark 1982, 1988; McBrearty and Brooks 2000). Conventionally, the MSA is cited as beginning by ca. 280 ka (Douze and Delagnes 2016) and ending by ca. 40 ka (Ambrose 1998). Recent scholarship, however, has demonstrated that stone reduction strategies typical of the MSA are present beyond this time range, with earlier occurrences at ~ 320 ka in the East African Rift (Brooks et al. 2018), in association with Homo sapiens fossils ca. 315 ka in the Maghreb (Richter et al. 2017), and persisting into the terminal Pleistocene or even Holocene of West Africa and the Horn of Africa (Scerri et al. 2017; Tribolo et al. 2017).
The early presence of Later Stone Age (LSA) sites in both East Africa (Ambrose 1998) and South Africa (d’Errico et al. 2012; Villa et al. 2012) by ~ 40 ka highlights that shifts in foraging technologies were not uniform with respect to geography or chronology. Research, particularly over the last 20 years, has sharpened the focus on the MSA as a period marked by innovation in both stone tool manufacture and other aspects of life that leave material traces. These include symbolic use of ochre and other materials (Henshilwood and Dubreuil 2011), regular big-game hunting using long-distance projectiles (O’Driscoll and Thompson 2018; Sahle et al. 2013), and ultimately as the period in which Homo sapiens dispersed out of the African continent between ca. 80 and 40 ka (Pagani et al. 2016). These observations, however, are largely based on robust datasets from areas of the continent that have historically received significant research attention: the southern tip of Africa, the East African Rift Valley, the Nile Valley, and the Maghreb. Significant spatial gaps therefore exist across the continent (Thompson et al. 2018), resulting in a patchwork understanding of behavioral differences across an environmentally varied space. Further, evidence for complex innovations is not ubiquitous across the known MSA record, even at very late-occurring MSA sites. In order to broaden the known temporal, geographic, and environmental contexts of MSA assemblages, we report stone artifacts and environmental reconstruction from Chaminade I (CHA-I, 9.956° S, 33.892° E), an open-air, late MSA site in Karonga, northern Malawi. This locality, at the southern tip of the western arm of the African Rift Valley, is one of only a few with dated MSA assemblages in east-central Africa (Mercader et al. 2009, 2012; Thompson et al. 2018; Wright et al. 2017).
The site of CHA-I preserves the techno-behavioral signatures of a late MSA population in a region of sub-Saharan Africa that has seen little detailed investigation. As a midpoint between the biogeographic regions of southern and eastern Africa, the Malawi archaeological record provides opportunities to examine variation in technological behavior and environmental preferences in the MSA. It is also a rare data point in the tropical savannas that characterized south-central Africa during the Late Pleistocene. Much emphasis has been placed on the role of climate change in mediating MSA population dynamics (Basell 2008), and the degree to which localized environments may have been buffered against extreme climate change has been explored in a synthesis by Blome et al. (2012). However, confidently dated MSA occurrences currently are exemplified by only three archaeological localities from the entire region that today encompasses Malawi, Zambia, Mozambique, Zimbabwe, and the Democratic Republic of the Congo (Blome et al. 2012). In contrast, there are several nearby high-resolution paleoclimatic datasets showing that central Africa experienced extreme fluctuations in precipitation, lake levels, and terrestrial environmental response during the Late Pleistocene (Garcin et al. 2006; Ivory et al. 2018; Johnson et al. 2016; Lyons et al. 2015; Tierney et al. 2008). This disjoin between the locations of dated and described archaeological data and the locations of rich paleoenvironmental records has made it difficult to assess larger patterns in the relationships between MSA people and their environments.
The primary aim of this analysis is to offer a full description of an in situ MSA site in Karonga, so that robust comparisons can be made to other archaeological sites. We do this in the following ways: (1) describe the technological character of the lithic assemblage, (2) examine the relationship between raw material use and reduction strategies, (3) reconstruct the depositional and post-depositional processes that have affected site formation, (4) report the ages of the archaeological material through the results of optically stimulated luminescence (OSL) analyses, (5) provide the environmental context of the CHA-I site, and (6) contextualize CHA-I in the regional setting of the local and continental MSA.
The African Middle Stone Age
The MSA as originally defined by Goodwin (1928) is a techno-cultural industrial designation marked by prepared core reduction, the elaboration of hunting tool forms, and the absence of stone tools characteristic of either the Earlier Stone Age (i.e., large core tools) or Later Stone Age (i.e., microliths). This definition continues to evolve as researchers learn more about variation within MSA technology, as well as its temporal and geographic boundaries (Growcutt and Blinkhorn 2013). Academic discourse on the MSA since Goodwin’s writing has largely focused on behavioral and biological evolutionary events and trends surrounding the emergence and spread of Homo sapiens, as it is generally accepted that the adaptive strategies acquired during this period enabled some, but not all, members of the Homo lineage to proliferate during the Middle and Later Pleistocene (Henn et al. 2011; Hublin and Klein 2011). In recent decades, MSA scholarship has centered around regionally specific reduction strategies and tools forms, evidence for increased dietary breadth, symbolic behavior, expanded foraging ranges, and in particular on specialized tool kits indicative of complex, extractive foraging strategies (Brown et al. 2012; Delagnes et al. 2016; Henshilwood and Dubreuil 2011; McCall and Thomas 2012; Sahle et al. 2014). These include the transport of raw material over considerable distances, standardization of tool forms, and evidence for composite or hafted technologies, which are strategies further expected to reflect some degree of population (i.e., cultural) cohesion or differentiation in response to regional environmental conditions and resource availability (Mackay et al. 2014).
Research in northern, eastern, and southern Africa has shown that the MSA, as a cultural/technological package, has developed discontinuously across the continent and over time (Johnson and McBrearty 2010; McBrearty and Brooks 2000; Sahle et al. 2014; Tribolo et al. 2015; Tryon and Faith 2013; Tryon et al. 2005; Van Peer et al. 2003; Wilkins and Chazan 2012), with technological hallmarks of MSA behavior both predating the minimum starting date of the MSA (~ 280 ka), and persisting in some areas into otherwise Later Stone Age periods (as recently as ~ 7.5 ka in the Horn of Africa; see Scerri et al. 2017; Tribolo et al. 2017). The diversity of behavioral repertoires evidenced in MSA contexts across Africa therefore likely reflects localized adaptations to specific environmental and cultural stimuli, rather than monolithic behavioral trends; this makes it critical to develop an understanding of MSA behavior in diverse paleoenvironmental and geographic zones (Thompson et al. 2018).
Environmental and Landscape Setting
The Karonga District of northern Malawi (Fig. 1) is situated within the Zambezian Phytogeographic Zone, characterized today by miombo woodlands containing a mosaic of tall Brachystegia trees and variable distributions of savanna grasses, hydromorphic grasses, and evergreen taxa (Mercader et al. 2011; White 1983). The climate is of an equatorial savanna (Aw in the Köppen-Geiger climate classification system), with hot, rainy summers and cool, dry winters (Kottek et al. 2006). Because it lies on the southern margin of the Intertropical Convergence Zone (ITCZ), most of the Lake Malawi catchment has only a single rainy season (November to April or May). The district is comprised of gently undulating topography on the distal end of an extensive alluvial fan system, bounded to the west by highlands of the southern Rift Valley (~ 2500 masl) and ~ 100 km to the east by Lake Malawi (~ 450 masl). To the south, the Chiweta Escarpment forms another physical boundary. In the past, these landscape features may have constrained foraging patterns for MSA populations. Rivers and streams incise the landscape, draining east into the lake, which is currently at a high-stand level.
Throughout the Pleistocene, the Lake Malawi catchment has been subject to extreme environmental variability, as seen in “megadrought” periods that have reduced lake levels by as much as 95% relative to today, transforming the local environment and subjecting human populations to changing configurations of water, vegetation, and animal resources (Cohen et al. 2007; Ivory et al. 2012; Lyons et al. 2015). The lake began to recover from the most recent extreme lowstand ca. 100 ka, and following 80 ka, lake level variability was relatively minor compared to other catchments such as Tanganyika (Bonnefille and Chalié 2000; Tierney et al. 2008) and Masoko (Garcin et al. 2006; Vincens et al. 2007). Following arid conditions of Marine Isotope Stage (MIS) 5b (87–82 ka), the overall balance of moisture has trended toward warmer and moister conditions until the Holocene (Ivory et al. 2012). Especially since ca. 60 ka, Lake Malawi has experienced an unusually long period of hydrologic overfilling (Lyons et al. 2015). Changes in orbital precession have had a limited impact on the overall distribution of rainfall since MIS 4, which ended by 57 ka (Lyons et al. 2015). All dated MSA sites in the Karonga District fall within the time of lake recovery since the last major arid period, and the majority of dated assemblages occur between 50 and 20 ka (Thompson et al. 2018). It is not clear if this increased archaeological visibility is because of an increase in population size or because of depositional factors.
Although the lake core records provide catchment-scale data on precipitation and vegetation, analysis of historical records shows that local-scale precipitation variability across the basin is high (Jury and Gwazantini 2002; Kumboyo et al. 2014; Ngongondo et al. 2001; Sene et al. 2016). This difference between catchment-scale and local proxies is also apparent in the Pleistocene. MSA sites were primarily preserved by alluvial fans that buried the sites and fluvial systems that activated within these settings created riparian zones that attracted human activity (Thompson et al. 2018; Wright et al. 2014, 2017). When Wright et al. (2017) compared phytolith assemblages from the site of Chaminade-II (CHA-II; ~ 100 m northeast of CHA-I) to vegetation proxies in the Lake Malawi Drill Core, they found that more woodland versus grassland was represented in the site and drill core records, respectively.
Previous Research on the Malawian MSA
Study of the Malawian MSA was initiated by J.D. Clark and colleagues in the 1950s (Clark 1954, 1956), with most focus on Karonga taking place in the 1960s (Clark 1966, 1967, 1968, 1972; Clark et al. 1966, 1970). MSA research in the region was largely discontinued afterwards with the exception of geoarchaeological work at the site of Mwanganda’s Village by Kaufulu (1983, 1990). The Malawi Earlier-Middle Stone Age Project (MEMSAP) renewed study of northern Malawi in 2009 (Thompson et al. 2012), including a reinvestigation of the hypothesized Sangoan-era elephant butchery site at Mwanganda’s Village (Wright et al. 2014). In addition to controlled archaeological excavation, it included extensive survey and geological trenching of MSA-bearing deposits within the alluvial fan systems comprising the Chitimwe Beds (Thompson et al. 2013, 2014, 2018; Wright et al. 2014, 2017; Zipkin et al. 2015). The CHA-I excavation was excavated in 2011 as one of the controlled excavations, as a way to establish how deeply buried artifacts were within the fan, if they were in situ, and the nature of their environmental contexts. Although overview artifact data have been published (Thompson et al. 2018; Wright et al. 2017), this study represents the first full assemblage and environmental reconstruction to be reported from a single MSA site in northern Malawi.
In addition to demonstrating a focus on past human occupation of riparian sites, prior work in Karonga has revealed a pattern in the age distributions of those sites that places all of them in the Late Pleistocene and most post-dating ca. 50 ka. These data accord with recent archaeological work on the Mozambican side of Lake Malawi, in which MSA through LSA sites are found on fluvially incised terraces and buried within fans (Bicho et al. 2016; Gonçalves et al. 2016; Mercader et al. 2012). The Malawian material suggests that although there is technological diversity in the MSA lithic artifact assemblages, there is little evidence for technological “complexity,” often identified in MSA contexts by flaking or retouching methods for the production of specific tool forms, preferential raw material use or treatment, or hafted technology (Thompson et al. 2018). Detailed analysis of the CHA-I artifact assemblages allows these larger patterns in behavior to be examined at the site scale, and at the level of individual knapping events, revealing the underlying technological decision-making strategies of MSA people.
A 10 × 1-m excavation trench running north-south was first established and stepped into the brick pit (squares A1-A10) to assess what material might have been missing because of the brick pit excavation. With the exception of a single artifact from the middle of the long trench, all of the artifacts were recovered from the southern extent of the trench (squares A8, A9, and A10), below the base of the modern brick-clay excavation. The southern-most two meters of the trench were then extended 1 m to the east (squares B9 and B10) in order to maximize the buried artifact sample. Therefore, although the total extent of the excavation was 12 m2, and the upper layers yielded dating and paleoenvironmental data, only the southernmost 5 m2 penetrated deeply enough to produce the artifacts described here. In the southwestern-most square (A10), a ~ 1-m sondage was extended beyond the closing depth of the rest of the southern excavation area to establish if further artifact-bearing deposits were below.
Excavation was conducted according to natural stratigraphic boundaries, unless those boundaries exceeded 10 cm in maximum thickness. In those cases, the excavation proceeded in 10-cm arbitrary levels (spits). In the artifact-bearing squares of the excavation, the number of spits varied by square and depositional Unit, with a maximum of seven spits from the exposed surface of the brick pit to the base of the Unit; the artifactually sterile sondage in square A10 was excavated as a single thick level, as a geological trench. Lithic artifacts were piece-plotted during excavation using a Nikon C-series 5″ total station; all sediments were then sieved through 5 mm screens to recover artifacts that were not identified during excavation. Total station-derived datasets for plotted material were incorporated into a GIS (Geographic Information System) for high-resolution mapping and spatial analysis. Flaked stone represents the only artifact class recovered at CHA-I; organic materials such as bone or shell have not been recovered from the iron oxide-rich lateritic soils that characterize the surficial deposits of the region. Indicating successive wet and dry seasons, laterites are typically acidic and thus unlikely to preserve organic material. It is only in small pockets of more calcareous sediment in the Chitimwe Beds that there is any potential for organic preservation, and none have been found at CHA-I. Ochre has the potential to preserve and has been recovered from a site reported ~ 1 km to the north, excavated in the 1960s (Clark et al. 1970). Microcharcoal may preserve in sediment samples prepared for pollen extraction, but the remains of fires and hearths do not.
At CHA-I, OSL samples were selected from the Units immediately below, containing, and above the artifact concentrations, as well as from near the top of the alluvial sequence. Micromorphology samples were taken to capture the boundaries between macroscopically visible sedimentary Units. Loose sediment samples were taken for pollen and phytolith analysis in a column from the north profile of the excavation, although because it was a step trench, some sets were offset horizontally from the next set of samples. Samples were taken every 20 cm depth or at a visible change in stratigraphy, whichever came first (Fig. 2).
Optically stimulated luminescence (OSL) ages on quartz grains from CHA-I, Karonga, Malawi
Depth (m below surface)
Grain size (μm)
Equivalent dose (Gy)a
Cosmic dose (mGy/year)d
Dose rate (mGy/year)
OSL age (ka)e
14.37 ± 0.76
22.1 ± 3.2
1.6 ± 0.1
13.1 ± 0.1
1.41 ± 0.02
10 ± 3
0.19 ± 0.02
2.63 ± 0.13
5.5 ± 0.4
45.33 ± 1.93
14.3 ± 2.1
1.5 ± 0.1
10.4 ± 0.1
1.29 ± 0.02
10 ± 3
0.16 ± 0.02
2.19 ± 0.11
20.7 ± 1.5
75.82 ± 3.76
19.1 ± 2.7
0.9 ± 0.1
7.1 ± 0.1
1.34 ± 0.02
10 ± 3
0.14 ± 0.01
1.87 ± 0.09
40.6 ± 3.7
74.94 ± 4.28
23.6 ± 3.5
0.7 ± 0.1
6.0 ± 0.1
1.34 ± 0.02
15 ± 5
0.12 ± 0.01
1.73 ± 0.09
43.4 ± 4.1
The dose rate (Dr) is an estimate of the exposure of quartz grains to ionizing radiation with the decay of the U/Th decay series, 40K, and the contribution of cosmic rays. Concentrations of U, Th, and K were determined by inductively coupled plasma mass spectrometry (ICP-MS) by Activation Laboratory Ltd., Ancaster, Ontario, Canada. The cosmic ray dose was calculated based upon the sample’s location, elevation, and depth below the ground surface (Prescott and Hutton 1994).
Calculation of De by the single aliquot protocols was accomplished for 28 to 34 aliquots (Table 1). For all samples, 70 to 100% aliquots were used to define the final De distribution and age determination; aliquots were removed from analysis because the recycling ratio was not between 0.90 and 1.10, the zero dose was > 5% of the natural emissions, or the error in equivalent dose determination is > 10%. De distributions were log normal and exhibited a range of overdispersion values from 24 to 14% (Table 1). An overdispersion percentage of a De distribution is an estimate of the relative standard deviation from a central De value in context of a statistical estimate of errors (Galbraith and Roberts 2012; Galbraith et al. 1999). A zero overdispersion percentage indicates high internal consistency in De values with 95% of the De values within 2σ errors. Overdispersion values ≤ 25% are routinely assessed for small aliquots of quartz grains that are well solar reset, like eolian sands (Meier et al. 2013; Olley et al. 2004; Wright et al. 2011), and this value is considered a threshold metric for calculation of a De value using the central age model of Galbraith et al. (1999).
Four micromorphological samples were taken from the north wall of the step-trench, three of which were aligned with OSL samples to aid in the evaluation and integration of results of both analyses. Thin sections from two of the oriented blocks were studied following methodology details in the SOM Section 2. Micromorphological description included sediment microstructure, mineralogy and grain size of the coarse fraction (> 100 μm), characteristics of the fine fraction (< 100 μm), porosity, nature and degree of bioturbation, spatial relationships of components, redoximorphic features, organic materials, clay coatings, and other attributes following Courty et al. (1989) and Stoops (2003). This methodology enables the study of sedimentary constituents and processes of deposition, as well as post-depositional alterations of the sediments resulting from soil formation, bioturbation, or groundwater fluctuations.
Sixteen paired phytolith and palynology samples were taken from the stepped north wall of the excavation, and one sample was taken from the base of the sondage wall at the southern extent of the trench. Phytolith extraction followed protocols established in Mercader et al. (2011) and is described in the SOM Section 3. An average of 451 phytoliths was tallied per sample in adjacent, non-overlapping lines across the cover slip (20 × 40 mm). Poaceae short cells were counted to > 200 cells. System microscopy was conducted at × 40 magnification (Motic BA410E; image software: Images Plus).
Palynomorphs (predominantly pollen and spores) and microcharcoal (particles < 125 μm) were extracted using the method developed by van der Kaars (1991) and discussed in depth by Moss (2013); details of extraction are provided in the SOM Section 4. Following extraction, pollen concentrate materials were slide-mounted in glycerol for analysis. The pollen sum consisted of a minimum of 200 palynomorphs or two completely counted slides, while the microcharcoal analysis involved counting all black angular fragments > 5 μm across three evenly spaced transects across all samples. Pollen and charcoal concentrations were determined from counts of exotic Lycopodium marker spores, which were added as a tablet (i.e., at the start of the analysis) with a known concentration of L. clavatum (Stockmarr 1971; Wang et al. 1999).
The stone tool assemblage from CHA-I was analyzed on metric and descriptive characteristics, based on the data collection methods established for previous MEMSAP excavations (Thompson et al. 2012). All objects were classed according to basic characteristics including: raw material, grain size, weight, maximum dimension, relative amount of cortex preserved, and technological component (core, flake, unidentifiable debitage or shatter, etc.). The degree of edge-rounding on each artifact was also quantified (following Thompson et al. 2012) to assess the possibility of post-depositional fluvial transport or subaerial exposure. Divided into four classes, artifacts were described as follows: 0, no edge rounding (i.e., in fresh/sharp condition); 1, slight edge rounding visible to the naked eye; 2, edge rounding to the degree that smaller features such as conchoidal ripples were obscured, though general artifact morphology and reduction characteristics are evident; 3, edge rounding significant enough to obscure major reduction characteristics or to modify artifact morphology. Metric measurements on cores, flakes, and other debitage captured size and shape characteristics and qualitative description permitted analysis of reduction strategies and trends in raw material use. Retouching location and invasiveness (the relative depth of a retouch scar across the surface of an artifact) were recorded when found (following Clarkson 2002), as was the presence of specialized tool forms such as points or blades.
Radial (or discoidal) cores were defined by centripetal flaking around a perimeter and onto one or both opposed hemispheres. Throughout MSA literature, the terms “radial” and “discoid” are used in various ways: (1) to differentiate cross-sectional attributes of cores (radial cores have thick cross-sections, while discoidal cores have flat cross-sections; Willoughby 2009), (2) according to the preferences of the analyst (for example, McCall (2006) and Tryon et al. (2012) consistently use “discoidal,” while Pleurdeau (2006) and Mackay et al. (2014) use “radial”), (3) to group non-Levallois radially flaked cores (discoids) and centripetal Levallois cores under the organizational rubric of “radial” (Villa et al. 2005), or (4) more-or-less interchangeably (McBrearty and Brooks 2000). Levallois cores are centripetally flaked and are distinguished on the basis of an asymmetry of hemisphere volumes, preparation of the preferential flake platform, and shaping of the Levallois surface for the removal of a preferential flake (Boëda 1995).
All artifacts were assessed for conjoins (ancient, unintentional breaks that did not result from the flaking process) and refits (sets of subsequent flake removals or flake-and-core sets; the distinction follows Sisk and Shea 2009). Refitting allows for a more complete understanding of reduction methods, the equifinality of multiple flaking schemes, and of the relative use-life of reduced stones. Refitting also permits an assessment of the degree of post-depositional disturbance or winnowing to which an assemblage may have been subjected; a high incidence of refitting artifacts suggests that little post-depositional reworking has modified the assemblage, whereas highly disturbed contexts are unlikely to preserve high frequencies of refitting pieces (Schick 1987).
Artifact attribute data resulting from lithic analysis were joined with total-station derived spatial data into a GIS, using ESRI ArcMap 10.2; full artifact data, including coordinate data, are provided as a supplementary database. GIS analysis consisted of three nested approaches: (1) map the distribution of raw material types and reduction strategies, (2) map refitting and conjoining artifacts, and (3) identify non-random clusters of artifacts using the ArcGIS Optimized Hot Spot Analysis (OHSA) functionality, similar to the analysis conducted for the nearby site of CHA-II (Wright et al. 2017). The OHSA tool analyzes two-dimensional datasets for spatial clustering along a third attribute; in this case, elevation was used as a third attribute, meaning that the tool identifies data points that are closely grouped on x-, y-, and z-axes. The resulting data layer identifies artifacts with high statistical significance, at 99%, 95%, and 90% confidence intervals, for both high and low z (elevation) values, as well as areas of no significant clustering. Given the small excavation area and limited depth (~ 5–10 cm) and slope of the artifact concentrations, it was unnecessary to normalize the z value data, as it might be necessary for datasets with wide ranges in elevation that could obscure mid-elevation clustering.
Dating and Stratigraphy
CHA-I stratigraphic description, artifact distribution, and OSL results
OSL sample and results
Unit 5: sandy alluvial fan deposits/laminated fill
Poorly sorted, subangular to subrounded medium to coarse sand fining up to a fine to medium sand
Soft, loose, nonsticky, nonplastic
2.5YR5/8-d (bottom), 2.5YR5/6-d (top)
Many root and insect disturbances
Merging, continuous boundary
n = 3
Minimal scatter, artifacts out of context
UIC3468 = 5470 ± 430
“A” in figs.
Unit 4 (top): gray clay/“brick pit”
Poorly sorted medium to very coarse, angular sandy clay loam
Soft, slightly friable, sticky, very plastic
7.5YR4/4-w, 10YR5/3-w, 10YR6/3-d, 10YR7/4-d
cm-scale intraclasts of moderately well sorted fine to medium sand; 1–2% 2–5-mm Fe-Mn nodules; many root and insect disturbances
n = 0
Unit 4 (base): Fe-Mn nodules/alluvial fan deposits
Poorly sorted angular to subangular coarse to very coarse sand and fine gravel fining upwards to very fine gravel/coarse sand
Slightly hard, friable, non-sticky, non-plastic
2.5Y6/2-w, 10YR7/4-d; weak redoximorphic mottling (10R6/6)
Discontinuously distributed 5–20% 0.5–2-mm Fe-Mn nodules; many root and insect disturbances
n = 893
Upper Artifact Concentration; ~ 92% of artifacts show little or no weathering; several refit/ conjoin sets
UIC3471 = 20,660 ± 1500
“B” in figs.
Unit 3: cobbles in fine gravel/stream channels
Poorly sorted, sporadically imbricated semi-prismoidal to semi-discoidal, rounded to well-rounded very coarse pebbles and cobbles in a matrix of poorly sorted angular to subangular coarse to very coarse sand and fine gravels; massive
Hard, friable, non-sticky, non-plastic
2.5Y6/2-w, 10YR7/4-d; cm-scale weak redoximorphic mottles (10R4/8)
2–5% dispersed Fe-Mn nodules (0.5-2 mm) and concentrated just above cobbles
Abrupt boundary dipping to the south
n = 323
Lower Artifact Concentration; ~ 87% of artifacts show little to no weathering; several refit/ conjoin sets
UIC3470 = 40,560 ± 3230
“C” in figs.
Unit 2b: pebble gravel
Poorly sorted angular to subangular coarse to very coarse sand to fine gravel (matrix supported); massive
Hard, friable, non-sticky, non-plastic
2.5YR5/4-w, 10YR7/4-d; weak mm- to cm-scale redoximorphic mottles (2.5YR6/8
Few root disturbances
Merging, continuous boundary
n = 75
Artifacts more diffuse and slightly more weathered than units above; no refit/conjoin sets
Unit 2a: gleyed clayey gravel
Poorly sorted, angular to subangular very coarse sand to medium gravel with 5–10% coarse rounded gravel; massive
Hard, friable, non-sticky, non-plastic
2.5YR4/8-w, 5Y7/1-w, 10YR6/3-d; weak mm- to cm-scale redoximorphic mottles (2.5YR5/8)
Many root disturbances
Merging, continuous boundary
n = 47
Artifacts more diffuse than units above; variable degrees of weathering; no refit/conjoin sets
UIC3555 = 43,390 ± 4125
“D” in figs.
Unit 1: fine gravel
Very poorly sorted, angular to subangular very coarse sand to fine gravel (matrix supported); massive; cm-scale intraclasts of poorly sorted medium- to coarse sand
Hard, friable, non-sticky, non-plastic
5YR4/2-w, 2.5Y7/3-d; weak mm- to cm-scale redoximorphic mottles (2.5YR5/8)
Common root disturbances
n = 0
Alluvial fan deposits are documented based on sediment texture and sorting characteristics (SOM Section 2) beginning prior to ~ 20.7 ka (OSL sample B). This constrains the upper age of all artifact deposition at the site (except for the artifact recovered from the top of Unit 5 that was likely introduced during a period of more recent alluvial activity). An abrupt depositional boundary was identified within the second stage of site formation, between Units 4 and 5, after which more alluvial fan deposits aggraded until they ceased sometime after ~ 5.5 ka (OSL sample A). Lateritic soil formation ensued after the cessation of alluvial fan deposition and contributed via pedogenic weathering and bioturbation to homogenization of the profile in Units 4 and 5; bedding structures and distinct sedimentation zones that may have been present during the initial sedimentation of the site were disturbed and overprinted. Laterite (plinthite) formation is typical of the Chitimwe Bed alluvial fan deposits in the Karonga area and has been described for the nearby excavation of CHA-II (Wright et al. 2014, 2017).
Unit 3 is contained in the upper part of sample 1988 (Fig. 5e) and consists of very fine to fine gravels among which are subrounded iron-manganese (Fe-Mn) nodules of fine gravel size (Fig. 5e, g). The fine fraction consists of pale yellowish, black-speckled, limpid clay, similar to Unit 2b. Few bioturbation features (roots; Fig. 5c) disrupt the subangular blocky microstructure, and locally, the clay is arranged in a very weak grano-striated b-fabric. Similar to Unit 2b, this Unit contains compound clay coatings, which have been deformed by vertic activity. Redox masses are more diffused compared to Unit 2b.
Sixteen paired sediment samples from the stratigraphic sequence were processed for phytolith and palynology analysis (nos. 700-715), and one phytolith sample was analyzed at the base of the sondage wall in square A10 (no. 1518). In both sets of samples, no. 715 did not produce sufficient phytolith or pollen counts to be included in analysis; two further samples (nos. 706, 713) also did not produce enough pollen for analysis, though all samples did produce microcharcoal.
A total of 7649 phytoliths were counted throughout the sequence, with a mean phytolith tally of 451 per sample. Preservation was adequate for morphometric analysis and type identification, and a total of 28 morphotypes were classified. The complete dataset appears in SOM Table S2; Fig. 6a illustrates the relative frequency of phytolith types through the CHA-I sequence.
Overall, the sampled sequence is dominated by arboreal morphotypes from bark tissue (Fig. 6a). The mean percentage of woody phytoliths throughout the column is 89% (min 63.93% [sample no. 700]; max 97.54% [sample no. 704]), reflecting a significant and consistent woody plant component around CHA-I that gradually increases through the lower, artifact-bearing Units. Grass clades are only minimally represented in the phytolith samples, typically at ~ 5% of overall phytolith totals. Variance in the relative frequency of woody and grass phytoliths support a division of the sequence into two main zones: an upper zone (sample nos. 700–703) in upper Unit 5, where woody phytoliths have a mean frequency of 77%, and a lower zone (sample nos. 704–714, 1518) running from lower Unit 5 to Unit 2b, in which woody phytoliths account for an average of 93% of the recovered samples. The upper zone marks an upward change toward tree cover contraction and spread of grasses. The most prominent change occurs in the youngest part of the column (sample nos. 700, 701) when cool-adapted pooids reach ~ 16.2%, and seasonally wet-adapted panicoids and xeric chloridoids also peak, albeit at low percentages; these increases are concomitant with a decrease in arboreal values. In the lower zone there are two arboreal hyper-abundance (> 90%) phases, at the middle of Unit 5 (sample no. 704) and Unit 3 (sample no. 714). The grasses from this phytolith zone, although uncommon, do reflect fluctuations with alternating higher and lower frequencies. They concurrently peak at sample no. 706 (mid-Unit 5) and sample nos. 709–711 (lower Unit 5/upper Unit 4) while the intervening values are relatively low. Samples 713 and 714 are associated with the majority of the MSA artifacts, and represent a time between ca. 40.6–20.7 ka when the relative abundance of arboreal phytoliths was at one of its highest points.
Sixteen palynology samples were taken in tandem with phytolith samples (nos. 700–715) but did not sample the base of the excavation/sondage. Three of the samples (706, 713, 715) did not produce sufficient pollen counts to be included in the analysis, though all samples did produce microcharcoal. Pollen from 32 taxa was recovered from the sequence, which we divide into seven groups: montane forest, evergreen forest, woodland taxa, herbaceous taxa, aquatic taxa, pteridophytes (ferns), and exotic taxa (Eucalyptus). Relative and absolute frequencies of recovered microcharcoal and pollen grains from the seven main groups are reported in Fig. 6b, and the full sequence of all 28 identified taxa are in SOM Figure S9.
In contrast to the phytolith samples where woody dicots are most common, herbs are the most commonly recovered pollen types throughout the sequence, accounting for ~ 15–50% of each sample (mean = ~ 35%). Arboreal taxa are the next most common, with an average frequency of ~ 20% through the sequence, followed by evergreen and montane taxa. Aquatic taxa and pteridophytes (ferns) are present throughout but show an increase in frequency in the upper half of the sequence. Absolute pollen and microcharcoal frequencies are low at the base of the excavation (Unit 2b), gradually increasing throughout Units 3 and 4 and into basal Unit 5, where frequencies again drop to barren levels and then increase to the surface. The changes in pollen and microcharcoal frequencies between samples indicate that bioturbation was limited to localized activity and did not homogenize the entire profile.
A total of 1341 lithic artifacts were recovered at CHA-I. The majority of these artifacts lie within either the “Upper Artifact Concentration” (n = 893) from the basal ~ 5 cm of Unit 4 or immediately below in the “Lower Artifact Concentration” (n = 323) at the top of Unit 3. Approximately 45% of the artifacts from these two concentrations are in fresh condition, and another 45% show only slight signs of mechanical or chemical weathering. Diffuse scatters of variably weathered artifacts were recovered in Units both above and below the two artifact concentrations (Table 2), though the very base of the excavation trench (Unit 1, in the square A10 sondage) was archaeologically sterile. We focus analysis of the lithic assemblages at CHA-I on the Upper and Lower Artifact Concentrations.
Like other Karonga MSA assemblages, the CHA-I assemblage is produced primarily on fine-grained quartzite and coarse-grained quartz; both materials derive from locally abundant river cobbles. Quartzite artifacts are more common in both the Upper and Lower Artifact Concentrations, at 62% and 78%, respectively. Artifacts made on chert and other fine-grained siliceous materials are present but rare in Units 4 and 3 (n = 6 and 1, respectively). Primary sources have not been identified in the Karonga region nor are these clasts common within sampled cobble beds.
A layer of small cobbles was excavated below the artifact concentrations (in unit 2b). Based on observations around CHA-I and other sites in Karonga, it is likely that cobble beds would have been exposed nearby, though it is not currently possible to determine an exact distance to such beds. The small cobbles from CHA-I were not collected at the time of excavation.
Upper Artifact Concentration
Lithic artifact assemblages composition, from the Upper Artifact Concentration (Unit 4) and Lower Artifact Concentration (Unit 3) at CHA-I
Upper Artifact Concentration
Lower Artifact Concentration
Subtotal (of assemblage)
At least three of the Levallois cores in the Upper Artifact Concentration were made on sizable primary flakes, which provided the natural convexity to the hemispheres variably exploited in Levallois reduction and minimized the degree of preparation on both surfaces for platform shaping or preferential removals. Radial and Levallois cores do not show significant differences in reduction intensity, as seen in the average number of flakes scars per core (~ 12) or the amount of cortex remaining (~ 0% and ~ 50–60% on opposed surfaces). However, radial cores are on average larger than Levallois cores (avg. L. 56.66 mm vs. 46.34 mm, respectively), suggesting that radial cores are slightly less reduced than Levallois cores relative to their size. The next most common core type is the “casual” or “minimally reduced” core, defined as a core with five or fewer flake scars. Because this definition is based on reduction intensity and not typology or technological organization, casual cores exhibit a range of reduction styles, and the difference between casual and other core types is often a matter of degree rather than kind. The casual cores in the Upper Artifact Concentration have an average length of 56.94 mm and most are reduced unifacially along a single platform edge (Fig. 7b), leaving a bulk of the cobble unmodified. Single platform core reduction is also typically focused on the smaller faces of more rectangular clasts while leaving the bulk of the cobble unmodified; the average single platform core has only six flake scars, retains ~ 67% cortex coverage, and is 56.5 mm in length. Multi-platform cores are more reduced than single platform cores, with an average of ten preserved flake scars and 30% remaining cortex coverage.
Complete flakes comprise only a small fraction of the Upper Artifact Concentration (n = 48, 46 of which are non-diagnostic). As with the cores from CHA-I, flakes show a variability of reduction styles and platform preparations, and the combination of these characteristics does not cluster in significant patterns; full data for complete flakes, including metric measurements, cortex coverage, and dorsal flake scar patterns and counts, are in SOM Table S4. In total, 22 of the 48 complete flakes (46%) retain cortex on the dorsal surface, with an average of 66.4% cortex coverage between them. Cortical platforms are the most common (29.2%), and five of the flakes are from primary core reduction. Plain and dihedral platforms (27.1% and 25%) are also common. Centripetal flake scars are the most common pattern (33.3%), followed by unidirectional flake scars (25%). Centripetal flakes preserve the greatest average number of dorsal scars (4.2) and the least amount of cortex (mean = 6.6% coverage), while unidirectional flakes have fewer dorsal scars (2.6) and greater cortex coverage (mean = 35%). This corresponds well to the core assemblage, where radially flaked cores are the most reduced and platform cores with parallel flake removals are among the least reduced. Only two flakes (one complete and one fragmentary) from the Upper Artifact Concentration bear the hallmarks of preferential Levallois flake reduction. This number likely underrepresents the true number of Levallois products, however, due to the equifinality of form between radial and Levallois core removals.
Other “specialized” forms, such as points or blades, are similarly rare within the Upper Artifact Concentration. Only two point-shaped flakes were recovered in the Upper 4, both of which were produced through unidirectional flaking. Given the paucity of this particular flake shape and the non-specific reduction method, we consider these “points” to be coincidental forms and not indicative of intentional point-production efforts. There is also no evidence for blade production at the site. The average length:width ratio for complete flakes in Unit 5 is 1.05 (σ = 0.28, CV = 27.0%), indicating that flakes are primarily short and broad and do not show great variance around this trend. Indeed, not a single flake from either Unit has a length:width ratio of 2:1, a discriminating characteristic of blades or “elongated” flakes. The most elongated flakes in the assemblage are those produced through platform reduction, featuring plain platforms and unidirectional flake scars (avg. L. 52.86 mm, avg. W. 37.76 mm, avg. L:W ratio 1.31). Modification of flake blanks into formalized tool types is also absent. Only three fragmentary retouched flakes were recovered, accounting for 0.4% of all artifacts Upper Artifact Concentration. The location and minimal intensity of retouch on these pieces does not appear to have been aimed at a specific, formalized tool type. No evidence of hafting residue or impact damage was noted in the assemblage. The Upper Artifact Concentration displays a high degree of fragmentation. Of the flake fraction, fragmentary flakes are nearly eight times more common than complete flakes, and angular shatter comprises ~ 47% of the total assemblage.
Lower Artifact Concentration
The Lower Artifact Concentration, although more diffuse than the assemblage above, has similar overall characteristics to the Upper Artifact Concentration. Cores are slightly less common (1.5% of total Lower assemblage, compared to 3.9% in the Upper assemblage) and slightly larger, though the length of each of the complete cores remains within one standard deviation of the mean length in the Upper concentration. Complete and fragmentary flakes are represented at similar proportions compared to the Upper Concentration. However, cortical platforms are much less common in the Lower concentration (only one of the 19 complete flakes [5%], versus 14 of 48 [29.2%] in the Upper assemblage); plain platforms account for 47% of the complete flakes, and the unidirectional flaking more common in the Upper concentration is only found on one of the 19 (5%) complete flakes. Cortex on flakes is also less common in the Lower Artifact Concentration: only four of the 19 complete flakes (21%) have cortex on the dorsal surface, with an average of 37.5% coverage between them.
As with the Upper concentration, the assemblage from the Lower Artifact Concentration lacks specialized or diagnostic tool forms: neither points, blades, nor retouched pieces were recovered, and only two fragmentary Levallois flakes were identified.
Refitting and Conjoining Artifacts
Two of the refitting groups are particularly worthy of note: refit groups 011 and 012 (Fig. 8) are both small quartz crystal cobbles (< 50 mm) that were likely flaked through initial hammer and anvil bipolar reduction, but suffered fracturing failures and ultimately shattered after two or three strikes each (Fig. 7d, f). These refit groups highlight several considerations that likely affected the reduction of various raw materials in this area: (1) while flaking quartz crystal can produce very sharp, durable cutting edges, flaking failure is fairly common, and may end in the complete obliteration of the toolstone clast and (2) the size and relative frequency of quartz crystal cobbles in the Chitimwe cobble beds (e.g., small and rare) results in minimal use of this particular material.
Artifact attributes such as raw material or reduction strategy did not map in spatially significant patterns. Quartz and quartzite artifacts are evenly distributed across excavation squares, and neither core nor flake characteristics appear in meaningful clusters. When mapped into the GIS, refit and conjoin groups cluster spatially and do not transgress across depositional layers. Several artifacts that comprise the refit and conjoin sets were recovered from the sieved artifact fraction and could only be mapped to the resolution of the excavation square. Nevertheless, most refitting and conjoining artifacts tend to occur within ~ 10–20 cm of each other.
Optimized Hot Spot Analysis (OHSA) was run on plotted artifacts from the two main artifact concentrations. Figure 9 shows the results of the OHSA for the Lower and Upper Artifact Concentrations, with clustering of artifacts around x-, y-, and z-axes for both low (blue) and high (red) elevations at three levels of statistical significance. The Lower Artifact Concentration (Unit 3) shows significant clustering at 90% and 95% confidence intervals in the northern area of the main excavation. A group of artifacts cluster at a 99% confidence interval in the southern extent of the excavation, where the largest number of refitting and conjoining artifacts in the Unit was also found. In the Upper Artifact Concentration (Unit 4), at least four areas of significant spatial clustering can be identified. In the southern excavation squares, artifacts cluster at both high and low relative elevations at 99% confidence, and refit groups often lie within these clusters.
When refit and conjoin data are mapped on to the OHSA results, refit groups occur most commonly in significantly associated artifact clusters. That sets of refitting and conjoining artifacts sit within non-random clusters of artifacts suggests that despite some post-depositional reworking and landform subsidence, discrete flaking and depositional events are preserved within the CHA-I assemblage.
The open-air setting and complex sedimentation and post-depositional history of CHA-I present challenges to interpreting how the site formed. Our reconstruction of the environmental setting and site formation processes combined field documentation with micromorphology, phytolith and palynology analysis, OSL dating, and the spatial analysis of recovered artifacts.
In the first depositional stage, the lowest Units of the site (Unit 1–3) document repeated stream alluvial deposition, particularly in Unit 2b (chronologically constrained between 40.6 ± 3.7 ka and 43.4 ± 4.1 ka). A lens of well-rounded pebbles caps Unit 2b, marking its boundary with Unit 3 (Fig. 5e). This pebble lens may represent the winnowed lag of a migrating channel, indicating the preservation of an exposed land surface at this position in the sequence. However, cut-and-fill processes or changes in depositional regimes could also explain the presence of these pebbles (Blair and McPherson 2009). Unit 2b is characterized by low porosity sediments with distinct redox masses and thick laminated (compound) clay coatings, reflecting partially saturated soil conditions alternating with drier periods during which clay illuviation could occur (Fig. 5e, f, h).
Unit 3 caps the first stage of deposition at CHA-I. The subrounded Fe-Mn nodules of Unit 3 (Figs. 5e, g) likely derive from surfaces eroded elsewhere and were subsequently redeposited together with the other very fine gravels of the Unit. As a result of their redeposition, the nodules are somewhat horizontally aligned Figs. 5e, g). The occurrence of reworked Fe-Mn nodules among the reworked gravels of Unit 3 reflects a change in sediment source as compared to the underlying Units. Distinct redox masses reach their maximum height in Unit 3, indicating that these lower strata have been impacted by fluctuating groundwater levels (sensu Vepraskas 1994) and are interpreted to have resulted in sediment compaction, landform subsidence, and a slight downward movement of sediment and associated artifacts. As redox features extend down the profile, there are no indications of persistent dry conditions in the lowermost three Units (Units 1–3) of the CHA-I sequence. Micromorphological observations indicate that both unit 2b and Unit 3 were still subject to seasonal fluctuations of the current water table. Woody dicot phytoliths and pollen from forest and woodland taxa are most common in the lower (older) depositional Units, and cobble beds present within and near the excavation area point to a wooded, riparian environment in which humans were making and discarding stone tools (Lower Artifact Concentration). Unit 3 ended with a break in sedimentation, after which a gravel surface was exposed for a length of time.
Although sparse artifact scatters exist throughout the lower portion of the sequence (Units 1-2b), and these show greater evidence of post-depositional weathering (and thus may be part of channel clast material), it is at the end of this depositional episode (in Unit 3) that substantial numbers of artifacts were deposited. A large component of the artifact assemblage is therefore likely in situ. After deposition of Unit 3 ceased, the Upper Artifact Concentration represents a scatter of lithic artifacts that likely accumulated on an exposed surface of gravels through manufacture and discard. This is indicated by the change in sedimentation between Units 3 and 4 from stream channel deposits to alluvial fan aggradation and the position of the artifacts at the base of Unit 4. Buried by accreting alluvial fan sediments prior to 20.7 ± 1.5 ka, these artifacts thus sit within Unit 4 but largely date to a preceding depositional hiatus between Units 3 and 4.
The existence of a former stable land surface somewhere above Unit 3 is evidenced by laminated (compound) clay coatings in both Units 2b and 3, as multiple phases of clay illuviation point to soil formation taking place over a prolonged period of time. The well-rounded, concentric Fe-Mn nodules found at the base of Unit 4 are endogenous pedogenic features (Fig. 5b, d), forming within the Unit 4 alluvium rather than being redeposited from elsewhere as were the Fe-Mn nodules in the Unit 3. Most likely, they started forming higher in the Unit. Angular sand inclusions of changing grain sizes in the concentric bands of the nodules point to a long history of formation during which the nodules moved downward and became concentrated at the base of the Unit. Factors contributing to their downward movement include shrink-and-swell processes, bioturbation, and the weight of the metallic Fe-Mn nodules.
It is difficult to infer the way phytoliths and pollen may have moved in the soil from the observation of large objects such as rounded Fe-Mn nodules, but it is possible that some of the lithic artifact assemblage in basal Unit 4 originally derived from somewhat higher in the Unit. This would date them to after the initiation of alluvial fan formation ca. 20.7 ka. However, the presence of refitting artifact groups in close spatial association, both vertical and horizontal, suggests that any movement of the artifact assemblage after initial deposition was minimal and that the majority of artifacts were recovered from a near-original depositional context, being buried by the sediment of Unit 4. The downward movement of coarse components of the sediment in Unit 4 may have also affected the pollen and phytolith records, though absolute counts in the samples from the Unit show that both pollen and phytoliths were recovered in greater numbers in the upper portion of the Unit. The pollen sample from the base of Unit 4 (sample no. 713) produced so few pollen grains that it was considered a barren sample. Had the downward movement of sediment in the Unit significantly repositioned pollen or phytoliths, they would likely also be concentrated at the base of the Unit, and exotic pollen (Eucalyptus) should not be restricted to only the top of the sequence.
Unit 4 has an increasingly open, better-aerated and mixed (bioturbated) microstructure, indicating that the vadose zone was reached (Fig. 5b–d). The sediments quickly fine upward from very coarse sand/very fine gravel with a large coarse fraction to a loamier coarse sand and fine gravel texture (Fig. 5b). At least part of the fining-up of the sediment is attributable to termite bioturbation, which is more pronounced in Unit 4 than in underlying sediments, and would have preferentially moved finer grained particles upward and coarse, heavy fragments downward (Crossley 1986).
Woody phytoliths remained dominant throughout the deposition of Unit 4, suggesting a relatively consistent tree cover in the area despite the possibility of taphonomic preservation bias (Albert et al. 2006; Garnier et al. 2012; Tsartsidou et al. 2007). We note variance between the phytolith and pollen records in regard to woody dicot vs. monocot grass cover at CHA-I and interpret this as reflecting different scales of influence, from site to catchment levels. The phytolith record is likely more relevant to the immediate area of the site while the pollen record samples the more regional environment. We also acknowledge the possibility of a taphonomic bias in terms of an underrepresentation of monocot grass phytoliths (Mercader et al. 2018). The combination of phytolith and pollen analyses highlights the vegetation mosaic typical of a miombo setting, with a strong signal of riparian and woody plants in the broader grassland environment. Nevertheless, the high frequency of herbs and woodland taxa pollen throughout the CHA-I sequence indicates primarily savanna-woodland conditions around the site as a result of either landscape clearance practices (e.g., burning, coppicing) or climate-induced vegetation change. Evergreen and montane forests are indicated in the pollen record at the sub-regional and regional scales, respectively. Higher rates of aquatic taxa and pteridophytes (ferns) toward the top of the sequence point to increasingly wetter conditions through the Holocene. Alluvial deposition at the site continued until ~ 5.5 ka, after which laterite soil formation intensified, and a reduction of tree cover is associated with the expansion of grass in the latest part of the sequence, mirroring modern conditions.
Stone Tool Systems
The lithic artifact assemblage from CHA-I reflects a flexible and generalized approach to stone tool production. Regional survey has demonstrated the abundance of toolstone throughout the Karonga region, and these locally available toolstone resources were targeted for stone tool making; to date, no artifacts have been recovered from MSA contexts around Karonga that have a clear exotic origin. The degree of cobble cortex and preserved flake scars on discarded cores shows that reduction efforts were not focused on extracting the maximum amount of flakes per unit of toolstone. In situ knapping events are documented at localities like CHA-I through refitting sets of artifacts, showing that MSA people were not visiting riparian areas solely to acquire cobbles for toolmaking. They spent time in active riparian systems, as shown by the arboreal phytolith signature in what catchment-scale data show was a mosaic and largely open environment with abundant grass (Ivory et al. 2018). The relatively high incidence of fragmentation found in the lithic artifact assemblage at CHA-I may further indicate that MSA foragers engaged in prolonged episodes of activity in these riparian zones, as trampling knapped stone has the documented effect of inducing significant breakage beyond that typical of knapping events (Driscoll et al. 2015; McBrearty et al. 1998).
Raw material frequencies of CHA-I Upper Artifact Concentration (Unit 4) and Lower Artifact Concentration (Unit 3), and sampled cobbles from nearby sites APS and BRU
Upper (Unit 4)
n · %
Lower (Unit 3)
n · %
≥ 20 mm cobbles
n · %
Mean max. dimension
≥ 20 mm cobbles
n · %
≥ 40 mm cobbles
n · %
Mean max. dimension
When taken with the greater frequency of quartzite artifacts in both Upper and Lower Artifact Concentrations at CHA-I, the implication is that cobble size is a first-stage determinant of raw material choice; there are fewer appropriately sized quartz cobbles from which to select, and therefore, quartzite becomes the material most likely to be used (Table 4). This finding further suggests that other aspects related to raw material choice (i.e., quality of material, predictability of fracture surfaces or edges, matrix homogeneity) may have had less demonstrable utility to the Karonga knappers, insofar as clasts appear to have been selected from cobble beds in relative proportion to their abundance above a minimum size. Quartz crystal cobbles sampled at BRU typically fall below 40 mm in maximum dimension. Thus, despite the advantages of quartz crystal in maintaining sharp cutting edges, it likely represented a relatively inefficient expenditure of time and energy when larger, more common, and more reliable toolstone clasts were available for the reduction strategies of the time. In spite of this, occasional attempts were made to exploit it even when there was apparently an abundance of other toolstone options in the immediate area. Upper size limits likely also influence raw material choice, because though very large (avg. 145 mm) quartzite cobbles exist in abundance across northern Malawi (Thompson et al. 2014), the majority of cores from CHA-I are less than half that size in maximum dimension, and retain sufficient cortex coverage to show that they had not been heavily reduced from a significantly greater size.
Platform and dorsal scar pattern attributes for complete flakes from the Upper and Lower Artifact Concentrations
Dorsal flake scar pattern
Upper Artifact Concentration (Unit 4)
Lower Artifact Concentration (Unit 3)
That platform cores are among the least reduced (avg. six scars) of the cores undercuts the idea that slightly elongated flakes were a targeted goal for most reduction efforts. Similarly, there is no obvious intentionality behind producing the squat flake blanks seen in the CHA-I assemblages. The flakes lack evidence for retouch needed to produce formalized “tool” types or for resharpening of cutting edges. Therefore, technological strategies at CHA-I are neither focused on maximization of toolstone clasts nor are they reflections of strict techno-cultural rules for the production of particularly shaped or sized flake blanks. A similar pattern was found during extensive surface survey (Thompson et al. 2014) and at nearby CHA-II (Wright et al. 2017). Apart from the factors of size and shape of raw material packages that may have influenced the use of certain raw materials or reduction methods over others, specific flaking methods do not appear linked to differential raw material use.
The most notable characteristic of the CHA-I stone tools is the rarity of formal or standardized flake production. Flakes in the CHA-I assemblage were not modified to facilitate hafting into compound or composite tools, though this does not preclude the hafting of unmodified flakes. The lack of specialized forms such as blades and points—the latter considered in most MSA contexts to have required hafting for intended and effective use (McBrearty and Brooks 2000)—may further suggest that multi-component technology was limited in the Karonga region. Retouched pieces are extremely rare, and specialized forms (such as points or blades) are non-existent at CHA-I. The absence of these tool forms from a late MSA context requires explanation.
One scenario might be that CHA-I samples a specific activity area that is not indicative of the whole suite of forager technologies in the Karonga area. Refitting artifacts in both the Upper and Lower Artifact Concentrations document discrete on-site knapping events, perhaps indicative of a workshop area situated to make best use of riparian resources. The relatively high incidence of Levallois cores to Levallois debitage may indicate on-site reduction and subsequent transport of flakes for use elsewhere; further supporting this idea is the fact that Levallois cores are more common at CHA-I (~ 25% of all cores) than at other Chaminade area sites (~ 5%) or sites further south (~ 6% in Sadala South study area; 0% in Ngara study area; Thompson et al. 2018). Alternatively, the apparent low frequency of Levallois flakes in the CHA-I assemblage may be a consequence of the minimal degree of preparation given to Levallois cores, as seen in the core sample recovered from CHA-I.
The lack of retouched or formalized tool types at CHA-I may alternatively reflect off-site modification of flake blanks. If this were the case, we would expect other sites in the area to produce typologically or technologically distinct lithic assemblages, and yet, the results of the lithic analysis at CHA-I are consistent with those from the Airport Site (Thompson et al. 2012), CHA-II (Wright et al. 2017), Mwanganda’s Village (Wright et al. 2014), and other sites in Karonga (Thompson et al. 2018). Throughout the Karonga region, formalized tool types, blades, and pointed forms are rare (Thompson et al. 2014, 2018), indicating that while the riparian resources at CHA-I likely provided an opportune locale for stone acquisition and reduction, the characteristics of the lithic assemblage reflect region-wide approaches to tool production and use. A broader survey of raw material use and core exploitation in the Karonga region holds to the trends witnessed at CHA-I, in particular the non-exhaustive use of locally available toolstone and technological flexibility in reduction strategies (Thompson et al. 2014). Taken together, these data show that stone tool manufacture at and around CHA-I in the later MSA of Malawi is not oriented toward strict technological end-goals. Rather, it seems the functional needs of late MSA foragers at Karonga were satisfied without complex or elaborated tool systems.
CHA-I in Larger Context
MSA foragers met their technological needs through varied means, in response to environmental and social conditions (Tryon and Faith 2016). Among these conditions are questions of hominin cognitive ability, population size and interaction, and subsistence risk. Thompson et al. (2018) reviewed these factors as they are predicted to affect forager technologies across the African MSA and found relative subsistence risk to offer the most compelling explanation for broader patterns observable across the continent. In this model, patchy or inconsistent resources are likely to result in a higher investment in technological systems, whereas predictable and abundant resources relax the need to develop complex technologies. The lithic artifact assemblages from CHA-I show little significant investment in complex technological design that is commonly found in other areas of Africa. In northeastern Africa, for example, points were predominantly produced through Nubian Levallois reduction methods, in which the shapes of flakes are predetermined through preparatory flaking of cores (Van Peer 1998). In northwestern Africa, the Aterian tradition documents the modification of flake blanks into tanged pieces, presumably to be inset in composite tools (Scerri 2013). The regular use of these strategies implies a link between specialized tool forms and specific activities—such as intercept hunting—within these forager populations. East African lithic assemblages often feature long-distance procurement of raw materials, blade production, bifacially retouched points, and early instances of microlithic implements (Ambrose 1998; Gliganic et al. 2012; Pleurdeau 2006; Tryon et al. 2012). The southern African record has produced a robust dataset of precocious time-constrained industries in which flake or blade blanks were retouched to create near-microlithic tool insets, again linked to specific environmental and demographic pressures (Brown et al. 2012; McCall 2007). In these examples, time and energy investments in stone tool technologies were oriented toward specific goals.
In the more immediate region of East-Central Africa, at the Songwe River Valley (Tanzania) (Willoughby 2001; Willoughby and Sipe 2002), and across Lake Malawi at the Mozambican sites of Ngalue Rockshelter, Mvumu, and other open-air localities in the Niassa Region (Bicho et al. 2016; Mercader et al. 2009, 2012), MSA assemblages are reported showing radial approaches to toolstone reduction, similar in style, frequency, and degree of exploitation to those seen at CHA-I. Notably, however, sites in both southern Tanzania and Mozambique also produced formalized, retouched tools such as scrapers, points, bifaces, and awls, in greater numbers than seen in any Karonga assemblages, as well as “heavy duty” tools such as choppers and core-axes, and grinding stones, all of which have been identified as expected components in the generalized MSA forager’s toolkit (Tryon and Faith 2013). All of these are absent from the CHA-I assemblage, and extensive survey and test-pitting have only identified them sporadically across the Karonga landscape (Thompson et al. 2018). From the perspective of the highly or even overdesigned toolkits in the northern and southern extremes of the continent, the lithic assemblages of Mozambique and southern Tanzania appear much more oriented toward a flexible foraging strategy, and the CHA-I assemblage extends this trend of diminishing technological investment.
The late persistence and expedient, unelaborated nature of the lithic technology at CHA-I is ostensibly at odds with a view of the MSA as a techno-chronological Unit that shows increasing behavioral and cognitive complexity following the emergence of Homo sapiens. Instead, it fits with an emerging disciplinary understanding of the many factors that influenced diversity in MSA forager behavior and technological choices. Compared to other regions across Africa from MIS4 to MIS2, northern Malawi enjoyed relatively stable and pluvial aggregate environmental conditions throughout its inhabitation by late MSA populations (Scholz et al. 2011). The datasets produced by excavation at CHA-I and nearby CHA-II (Wright et al. 2017) indicate the preservation of woody phytolith taxa during the LGM, suggesting perpetuation of a relatively moist climatic regime even as the northern portions of the continent faced severe aridification (Blome et al. 2012). Successive cycles of wetting and drying of the landscape, as documented through micromorphological analysis, further support this interpretation. The CHA-I excavation documents a riparian environment, where toolstone and food resources would have been regularly accessible; lithic raw materials were likely selected from the immediate environment, reduced and discarded on-site. In the context of broader patterns in the MSA, the late age of CHA-I shows that by ~ 40 ka, not all African populations were making substantial changes to their technological or behavioral patterns that archaeologists identify with the start of the LSA. Instead, in the relatively stable environments of the Lake Malawi basin, flexibility and expediency in lithic technology continued to provide the necessary means to meet daily foraging requirements.
We thank our collaborators at the Malawi Ministry of Civic Education, Culture, and Community Development for their assistance and permission in facilitating this research. The CHA-I field season was executed in 2011, and special thanks are given to Potiphar Kaliba (Deputy Director of Culture at the time), Oris Malijani (Senior Antiquities Officer), and an outstanding team of local crew—many of whom continue to work with us today. These include Moses Nyondo, Kondwani Mwafulirwa, Henry Kalinga, and Daudi Mwangomba. Flora Schilt thanks Prof. Dr. Christopher Miller and Dr. Susan Mentzer for their supervision, as well as Panagiotis Kritikakis for preparation of the thin sections for micromorphological analysis. We also thank the editors and six anonymous reviewers, whose thoughtful comments contributed substantially to the improvement of this manuscript.
Fieldwork and analysis were funded by National Geographic-Waitt Foundation grant W115-10 and the Australian Research Council Discovery Project DP110101305. Funding for the micromorphological analysis was provided by the Deutsche Forschungsgemeinschaft (MI 1748/3-1, MI 1748/1-1, and ME 4406/1-1).
Compliance with Ethical Standards
Conflict of Interest
The authors declare that there is no conflict of interest.
- Albert, R. M., Bamford, M. K., & Cabanes, D. (2006). Taphonomy of phytoliths and macroplants in different soils from Olduvai Gorge (Tanzania) and the application to Plio-Pleistocene palaeoanthropological samples. Quaternary International, 148, 78–94. https://doi.org/10.1016/j.quaint.2005.11.026.CrossRefGoogle Scholar
- Bicho, N., Haws, J., Raja, M., Madime, O., Gonçalves, C., Cascalheira, J., Benedetti, M., Pereira, T., & Aldeias, V. (2016). Middle and Late Stone Age of the Niassa region, northern Mozambique. Preliminary results. Quaternary International, 404, 87–99. https://doi.org/10.1016/j.quaint.2015.09.059.CrossRefGoogle Scholar
- Blome, M. W., Cohen, A. S., Tryon, C. A., Brooks, A. S., & Russell, J. (2012). The environmental context for the origins of modern human diversity: a synthesis of regional variability in African climate 150,000-30,000 years ago. Journal of Human Evolution, 62(5), 563–592. https://doi.org/10.1016/j.jhevol.2012.01.011.CrossRefGoogle Scholar
- Boëda, E. (1995). Levallois: a volumetric construction, methods, a technique. In The definition and interpretation of Levallois technology. (Vol. 23, pp. 41-65): Prehistory Press Madison.Google Scholar
- Brooks, A. S., Yellen, J. E., Potts, R., Behrensmeyer, A. K., Deino, A. L., Leslie, D. E., Ambrose, S. H., Ferguson, J. R., d’Errico, F., Zipkin, A. M., Whittaker, S., Post, J., Veatch, E. G., Foecke, K., & Clark, J. B. (2018). Long-distance stone transport and pigment use in the earliest Middle Stone Age. Science, 360(6384), 90–94.CrossRefGoogle Scholar
- Brown, K. S., Marean, C. W., Jacobs, Z., Schoville, B. J., Oestmo, S., Fisher, E. C., Bernatchez, J., Karkanas, P., & Matthews, T. (2012). An early and enduring advanced technology originating 71,000 years ago in South Africa. Nature, 491(7425), 590–593. https://doi.org/10.1038/nature11660.CrossRefGoogle Scholar
- Clark, J. D. (1954). Upper Sangoan industries from northern Nyasaland and the Luangwa Valley: a case of environmental differentiation? South African Journal of Science, 50(8), 201–208 http://hdl.handle.net/10520/AJA00382353_2810.Google Scholar
- Clark, J. D. (1956). Prehistory in Nyasaland. The Nyasaland Journal, 9(1), 92–119.Google Scholar
- Clark, J. D. (1966, The Society of Malawi Journal). Archaeology in Malawi., 19(2), 15–25 www.jstor.org/stable/29778138.
- Clark, J. D. (1968). Antiquities programme: Malawi - (mission) 19 July-15 September 1968. UNESCO Document (pp. 73). Paris.Google Scholar
- Clark, J. D. (1972). Archaeological mission: Malawi - (mission) July-September 1972. UNESCO Document (pp. 17). Paris.Google Scholar
- Clark, J. D., Haynes, C. V., Mawby, J. E., & Gautier, A. (1970). Interim report on palaeoanthropological investigations in the Lake Malawi Rift. Quaternaria, 13, 305–354.Google Scholar
- Clark, J. D., Stephens, E. A., & Coryndon, S. C. (1966). Pleistocene fossiliferous lake beds of the Malawi (Nyasa) Rift: a preliminary report. American Anthropologist, 68(2, Part 2: Recent Studies in Paleoanthropology), 46-87, https://doi.org/10.1525/aa.1966.68.2.02a00960.
- Cohen, A., Stone, J. R., Beuning, K. R. M., Park, L. E., Rainthal, P. N., Dettman, D., Scholz, C. A., Johnson, T. C., King, J. W., Talbot, M. R., Brown, E. T., & Ivory, S. J. (2007). Ecological consequences of early Late Pleistocene megadroughts in tropical Africa. Proceedings of the National Academy of Science, 104, 16422–16427. https://doi.org/10.1073/pnas.0703873104.CrossRefGoogle Scholar
- Courty, M. A., Goldberg, P., & Macphail, R. I. (1989). Soils and micromorphology in archaeology. Cambridge: Cambridge University Press.Google Scholar
- Delagnes, A., Schmidt, P., Douze, K., Wurz, S., Bellot-Gurlet, L., Conard, N. J., Nickel, K. G., Van Niekerk, K. L., & Henshilwood, C. S. (2016). Early evidence for the extensive heat treatment of Silcrete in the Howiesons Poort at Klipdrift Shelter (layer PBD, 65 ka), South Africa. PLoS One, 11(10), e0163874. https://doi.org/10.1371/journal.pone.0163874.CrossRefGoogle Scholar
- Galbraith, R. F., Roberts, R. G., Laslett, G. M., Yoshida, H., & Olley, J. M. (1999). Optical dating of single and multiple grains of quartz from Jinmium rock shelter, northern Australia, part 1, experimental design and statistical models. Archaeometry, 41, 339–364. https://doi.org/10.1111/j.1475-4754.1999.tb00987.x.CrossRefGoogle Scholar
- Garcin, Y., Williamson, D., Taieb, M., Vincens, A., Mathé, P. E., & Majule, A. (2006). Centennial to millennial changes in maar-lake deposition during the last 45,000 years in tropical Southern Africa (Lake Masoko, Tanzania). Palaeogeography, Palaeoclimatology, Palaeoecology, 239(3), 334–354. https://doi.org/10.1016/j.palaeo.2006.02.002.CrossRefGoogle Scholar
- Gliganic, L. A., Jacobs, Z., Roberts, R. G., Dominguez-Rodrigo, M., & Mabulla, A. Z. (2012). New ages for Middle and Later Stone Age deposits at Mumba rockshelter, Tanzania: optically stimulated luminescence dating of quartz and feldspar grains. Journal of Human Evolution, 62(4), 533–547. https://doi.org/10.1016/j.jhevol.2012.02.004.CrossRefGoogle Scholar
- Goodwin, A. J. H. (1928). An introduction to the Middle Stone Age in South Africa. South African Journal of Science, 25, 410–418.Google Scholar
- Henn, B. M., Bustamante, C. D., Mountain, J. L., & Feldman, M. W. (2011). Reply to Hublin and Klein: locating a geographic point of dispersion in Africa for contemporary humans. Proceedings of the National Academy of Sciences, 108(28), E278–E278. https://doi.org/10.1073/pnas.1107300108.CrossRefGoogle Scholar
- Henshilwood, C. S., van Niekerk, K. L., Wurz, S., Delagnes, A., Armitage, S. J., Rifkin, R. F., Douze, K., Keene, P., Haaland, M. M., Reynard, J., Discamps, E., & Mienies, S. S. (2014). Klipdrift Shelter, southern Cape, South Africa: preliminary report on the Howiesons Poort layers. Journal of Archaeological Science, 45, 284–303. https://doi.org/10.1016/j.jas.2014.01.033.CrossRefGoogle Scholar
- Ivory, S. J., Lézine, A.-M., Vincens, A., & Cohen, A. S. (2012). Effect of aridity and rainfall seasonality on vegetation in the southern tropics of East Africa during the Pleistocene/Holocene transition. Quaternary Research, 77(1), 77–86. https://doi.org/10.1016/j.yqres.2011.11.005.CrossRefGoogle Scholar
- Johnson, T. C., Werne, J. P., Brown, E. T., Abbott, A., Berke, M., Steinman, B. A., Halbur, J., Contreras, S., Grosshuesch, S., Deino, A., & Scholz, C. A. (2016). A progressively wetter climate in southern East Africa over the past 1.3 million years. Nature, 537(7619), 220–224. https://doi.org/10.1038/nature19065.CrossRefGoogle Scholar
- Kaufulu, Z. M. (1983). The geological context of some early archaeological sties in Kenya, Malawi, and Tanzania: microstratigraphy, site formation and interpretation. Berkeley: University of California.Google Scholar
- Lyons, R. P., Scholz, C. A., Cohen, A. S., King, J. W., Brown, E. T., Ivory, S. J., Johnson, T. C., Deino, A. L., Reinthal, P. N., & McGlue, M. M. (2015). Continuous 1.3-million-year record of East African hydroclimate, and implications for patterns of evolution and biodiversity. Proceedings of the National Academy of Sciences, 201512864, https://doi.org/10.1073/pnas.1512864112.
- Meier, H. A., Nordt, L. C., Forman, S. L., & Dreise, S. G. (2013). Late Quaternary alluvial history of the middle Owl Creek drainage basin in central Texas: a record of geomorphic response to environmental change. Quaternary International, 306, 24–41. https://doi.org/10.1016/j.quaint.2013.07.010.CrossRefGoogle Scholar
- Mercader, J., Akeju, T., Brown, M., Bundala, M., Collins, M. J., Copeland, L., Crowther, A., Dunfield, P., Henry, A., Inwood, J., Itambu, M., Kim, J.-J., Larter, S., Longo, L., Oldenburg, T., Patalano, R., Sammynaiken, R., Soto, M., Tyler, R., & Xhauflair, H. (2018). Exaggerated expectations in ancient starch research and the need for new taphonomic and authenticity criteria. Facets, 3, 777–798. https://doi.org/10.1139/facets-2017-0126.CrossRefGoogle Scholar
- Pagani, L., Lawson, D. J., Jagoda, E., Mörseburg, A., Eriksson, A., Mitt, M., Clemente, F., Hudjashov, G., DeGiorgio, M., Saag, L., Wall, J. D., Cardona, A., Mägi, R., Sayres, M. A. W., Kaewert, S., Inchley, C., Scheib, C. L., Järve, M., Karmin, M., Jacobs, G. S., Antao, T., Iliescu, F. M., Kushniarevich, A., Ayub, Q., Tyler-Smith, C., Xue, Y., Yunusbayev, B., Tambets, K., Mallick, C. B., Saag, L., Pocheshkhova, E., Andriadze, G., Muller, C., Westaway, M. C., Lambert, D. M., Zoraqi, G., Turdikulova, S., Dalimova, D., Sabitov, Z., Sultana, G. N. N., Lachance, J., Tishkoff, S., Momynaliev, K., Isakova, J., Damba, L. D., Gubina, M., Nymadawa, P., Evseeva, I., Atramentova, L., Utevska, O., Ricaut, F.-X., Brucato, N., Sudoyo, H., Letellier, T., Cox, M. P., Barashkov, N. A., Skaro, V., Mulahasanovic, L., Primorac, D., Sahakyan, H., Mormina, M., Eichstaedt, C. A., Lichman, D. V., Abdullah, S., Chaubey, G., Wee, J. T. S., Mihailov, E., Karunas, A., Litvinov, S., Khusainova, R., Ekomasova, N., Akhmetova, V., Khidiyatova, I., Marjanović, D., Yepiskoposyan, L., Behar, D. M., Balanovska, E., Metspalu, A., Derenko, M., Malyarchuk, B., Voevoda, M., Fedorova, S. A., Osipova, L. P., Lahr, M. M., Gerbault, P., Leavesley, M., Migliano, A. B., Petraglia, M., Balanovsky, O., Khusnutdinova, E. K., Metspalu, E., Thomas, M. G., Manica, A., Nielsen, R., Villems, R., Willerslev, E., Kivisild, T., & Metspalu, M. (2016). Genomic analyses inform on migration events during the peopling of Eurasia. [Letter]. Nature, 538(7624), 238. https://doi.org/10.1038/nature19792.CrossRefGoogle Scholar
- Richter, D., Grün, R., Joannes-Boyau, R., Steele, T. E., Amani, F., Rué, M., Fernandes, P., Raynal, J.-P., Geraads, D., Ben-Ncer, A., Hublin, J.-J., & McPherron, S. P. (2017). The age of the hominin fossils from Jebel Irhoud, Morocco, and the origins of the Middle Stone Age. Nature, 546, 293–296. https://doi.org/10.1038/nature22335.CrossRefGoogle Scholar
- Sahle, Y., Morgan, L. E., Braun, D. R., Atnafu, B., & Hutchings, W. K. (2014). Chronological and behavioral contexts of the earliest Middle Stone Age in the Gademotta Formation, Main Ethiopian Rift. Quaternary International, 331, 6–19. https://doi.org/10.1016/j.quaint.2013.03.010.CrossRefGoogle Scholar
- Scerri, E. M., Blinkhorn, J., Niang, K., Bateman, M. D., & Groucutt, H. S. (2017). Persistence of Middle Stone Age technology to the Pleistocene/Holocene transition supports a complex hominin evolutionary scenario in West Africa. Journal of Archaeological Science: Reports, 11, 639–646. https://doi.org/10.1016/j.jasrep.2017.01.003.CrossRefGoogle Scholar
- Schick, K. D. (1987). Experimentally-derived criteria for assessing hydrologic disturbance of archaeological sites. Natural formation processes and the archaeological record, 352, 86–107.Google Scholar
- Scholz, C. A., Cohen, A. S., Johnson, T. C., King, J., Talbot, M. R., & Brown, E. T. (2011). Scientific drilling in the Great Rift Valley: the 2005 Lake Malawi Scientific Drilling Project—an overview of the past 145,000 years of climate variability in Southern Hemisphere East Africa. Palaeogeography, Palaeoclimatology, Palaeoecology, 303, 3–19. https://doi.org/10.1016/j.palaeo.2010.10.030.CrossRefGoogle Scholar
- Stockmarr, J. (1971). Tablets with spores used in absolute pollen analysis. Pollen et Spores, 13, 615–621.Google Scholar
- Stoops, G. (2003). Guidelines for analysis and description of soil and regolith thin sections. Madison, Wisconsin: Soil Science Society of America, Inc..Google Scholar
- Thompson, E., Williams, H. M., & Minichillo, T. (2010). Middle and late Pleistocene Middle Stone Age lithic technology from Pinnacle Point 13B (Mossel Bay, Western Cape Province, South Africa). Journal of Human Evolution, 59(3–4), 358–377. https://doi.org/10.1016/j.jhevol.2010.07.009.CrossRefGoogle Scholar
- Thompson, J. C., Mackay, A., de Moor, V., & Gomani-Chindebvu, E. (2014). Catchment survey in the Karonga District: a landscape-scale analysis of provision and core reduction strategies during the Middle Stone Age of northern Malawi. African Archaeological Review, 31, 447–478. https://doi.org/10.1007/s10437-014-9167-2.CrossRefGoogle Scholar
- Thompson, J. C., Mackay, A., Nightingale, S., Wright, D. K., Choi, J.-H., Blackmore, H., Welling, M., & Gomani-Chindebvu, E. (2018). Ecological risk, demography, and geographic patterning of technological complexity in the Late Pleistocene of Africa. Journal of Quaternary Science, 33(3), 261–284. https://doi.org/10.1002/jqs.3002.CrossRefGoogle Scholar
- Tribolo, C., Asrat, A., Bahain, J.-J., Chapon, C., Douville, E., Fragnol, C., Hernandez, M., Hovers, E., Leplongeon, A., & Martin, L. (2017). Across the gap: geochronological and sedimentological analyses from the Late Pleistocene-Holocene sequence of Goda Buticha, Southeastern Ethiopia. PLoS One, 12(1), e0169418. https://doi.org/10.1371/journal.pone.0169418.CrossRefGoogle Scholar
- Tribolo, C., Rasse, M., Soriano, S., & Huysecom, E. (2015). Defining a chronological framework for the Middle Stone Age in West Africa: comparison of methods and models for OSL ages at Ounjougou (Mali). Quaternary Geochronology, 29, 80–96. https://doi.org/10.1016/j.quageo.2015.05.013.CrossRefGoogle Scholar
- Tryon C. A., & Faith J. T. (2016) A demographic perspective on the Middle to Later Stone Age transition from Nasera rockshelter, Tanzania. Phil. Trans. R. Soc. B 371:20150238.Google Scholar
- Tryon, C. A., Peppe, D. J., Tyler Faith, J., Van Plantinga, A., Nightingale, S., Ogondo, J., & Fox, D. L. (2012). Late Pleistocene artefacts and fauna from Rusinga and Mfangano islands, Lake Victoria, Kenya. Azania: Archaeological Research in Africa, 47(1), 14–38. https://doi.org/10.1080/0067270x.2011.647946.CrossRefGoogle Scholar
- Tsartsidou, G., Lev-Yadun, S., Albert, R. M., Miller-Rosen, A., Efstratiou, N., & Weiner, S. (2007). The phytolith archaeological record: strengths and weaknesses evaluated based on a quantitative modern reference collection from Greece. Journal of Archaeological Science, 34(8), 1262–1275.CrossRefGoogle Scholar
- Van Peer, P. (1998). The Nile Corridor and the Out-of-Africa model: an examination of the archaeological record. Current Anthropology, 39(S1, Special Issue: The Neanderthal Problem and the Evolution of Human Behavior.), S115-S140, doi: https://doi.org/10.1086/204692.
- Van Peer, P., Fullagar, R., Stokes, S., Bailey, R. M., Moeyersons, J., Steenhoudt, F., Geerts, A., Vanderbeken, T., De Dapper, M., & Geus, F. (2003). The Early to Middle Stone Age transition and the emergence of modern human behaviour at site 8-B-11, Sai Island, Sudan. Journal of Human Evolution, 45(2), 187–193. https://doi.org/10.1016/s0047-2484(03)00103-9.CrossRefGoogle Scholar
- Vepraskas, M. J. (1994). Redoximorphic features for identifying aquic conditions. North Carolina Agricultural Research Service, Raleight, North Carolina., Technical Bulletin, 301, 33.Google Scholar
- Vepraskas, M. J., Wilding, L. P., & Drees, L. R. (1993). Aquic conditions for soil taxonomy: concepts, soil morphology and micromorphology. In A. J. Ringrose-Voase & G. S. Humphreys (Eds.), Developments in soil science (Vol. 22, pp. 117–131). Amsterdam: Elsevier.Google Scholar
- Villa, P., Soriano, S., Tsanova, T., Degano, I., Higham, T. F. G., d’Errico, F., Backwell, L., Lucejko, J. J., Colombini, M. P., & Beaumont, P. B. (2012). Border Cave the the beginning of the Later Stone Age in South Africa. Proceedings of the National Academy of Science, 109(33), 13208–13213.CrossRefGoogle Scholar
- Wang, X., van der Kaars, S., Kershaw, P., Bird, M., & Jansen, F. (1999). A record of fire, vegetation and climate through the last three glacial cycles from Lombok Ridge core G6-4, eastern Indian Ocean, Indonesia. Palaeogeography, Palaeoclimatology, Palaeoecology, 147, 241–256. https://doi.org/10.1016/S0031-0182(98)00169-2.CrossRefGoogle Scholar
- White, F. (1983). The vegetation of Africa: a descriptive memoir to accompany the UNESCO/AETFAT/UNSO vegetation map of Africa. Natural Resources Research Report XX, UNESCO. Paris.Google Scholar
- Wilkins, J., & Chazan, M. (2012). Blade production 500 thousand years ago at Kathu Pan 1, South Africa: support for a multiple origins hypothesis for early Middle Pleistocene blade technologies. Journal of Archaeological Science, 39(6), 1883–1900. https://doi.org/10.1016/j.jas.2012.01.031.CrossRefGoogle Scholar
- Wright, D. K., Thompson, J. C., Mackay, A., Welling, M., Forman, S. L., Price, G., Zhao, J.-x., Cohen, A. S., Malijani, O., & Gomani-Chindebvu, E. (2014). Renewed geoarchaeological investigations of Mwanganda’s Village (Elephant Butchery Site), Karonga, Malawi. Geoarchaeology, 29, 98–120. https://doi.org/10.1002/gea.21469.CrossRefGoogle Scholar
- Wright, D. K., Thompson, J. C., Schilt, F., Cohen, A. S., Choi, J.-H., Mercader, J., Nightingale, S., Miller, C. E., Mentzer, S. M., Walde, D., Welling, M., & Gomani-Chindebvu, E. (2017). Approaches to Middle Stone Age landscape archaeology in tropical Africa. Journal of Archaeological Science, 77, 64–77. https://doi.org/10.1016/j.jas.2016.01.014.CrossRefGoogle Scholar
- Zipkin, A. M., Hanchar, J. M., Brooks, A. S., Grabowski, M. W., Thompson, J. C., & Gomani-Chindebvu, E. (2015). Ochre fingerprints: distinguishing among Malawian mineral pigment sources with Homogenized Ochre Chip LA-ICPMS. Archaeometry, 57(2), 297–317. https://doi.org/10.1111/arcm.12090.CrossRefGoogle Scholar