Multiple lines of genetic and archaeological evidence suggest that there were major demographic changes in the terminal Late Pleistocene epoch and early Holocene epoch of sub-Saharan Africa1–4. Inferences about this period are challenging to make because demographic shifts in the past 5,000 years have obscured the structures of more ancient populations3,5. Here we present genome-wide ancient DNA data for six individuals from eastern and south-central Africa spanning the past approximately 18,000 years (doubling the time depth of sub-Saharan African ancient DNA), increase the data quality for 15 previously published ancient individuals and analyse these alongside data from 13 other published ancient individuals. The ancestry of the individuals in our study area can be modelled as a geographically structured mixture of three highly divergent source populations, probably reflecting Pleistocene interactions around 80–20 thousand years ago, including deeply diverged eastern and southern African lineages, plus a previously unappreciated ubiquitous distribution of ancestry that occurs in highest proportion today in central African rainforest hunter-gatherers. Once established, this structure remained highly stable, with limited long-range gene flow. These results provide a new line of genetic evidence in support of hypotheses that have emerged from archaeological analyses but remain contested, suggesting increasing regionalization at the end of the Pleistocene epoch.

The archaeology of East Africa during the last ~65,000 years plays a central role in debates about the origins and dispersal of modern humans, Homo sapiens. Despite the historical importance of the region to these discussions, reliable chronologies for the nature, tempo, and timing of human behavioral changes seen among Middle Stone Age (MSA) and Later Stone Age (LSA) archaeological assemblages are sparse. The Kisese II rockshelter in the Kondoa region of Tanzania, originally excavated in 1956, preserves a ≥ 6-m-thick archaeological succession that spans the MSA/LSA transition, with lithic artifacts such as Levallois and bladelet cores and backed microliths, the recurrent use of red ochre, and >5,000 ostrich eggshell beads and bead fragments. Twenty-nine radiocarbon dates on ostrich eggshell carbonate make Kisese II one of the most robust chronological sequences for understanding archaeological change over the last ~47,000 years in East Africa. In particular, ostrich eggshell beads and backed microliths appear by 46–42 ka cal BP and occur throughout overlying Late Pleistocene and Holocene strata. Changes in lithic technology suggest an MSA/LSA transition that began 39–34.3 ka, with typical LSA technologies in place by the Last Glacial Maximum. The timing of these changes demonstrates the time-transgressive nature of behavioral innovations often linked to the origins of modern humans, even within a single region of Africa.

There are multiple hypotheses for human responses to glacial cycling in the Late Pleistocene, including changes in population size, interconnectedness, and mobility. Lithic technological analysis informs us of human responses to environmental change because lithic assemblage characteristics are a reflection of raw material transport, reduction, and discard behaviors that depend on hunter-gatherer social and economic decisions. Pinnacle Point Site 5–6 (PP5-6), Western Cape, South Africa is an ideal locality for examining the influence of glacial cycling on early modern human behaviors because it preserves a long sequence spanning marine isotope stages (MIS) 5, 4, and 3 and is associated with robust records of paleoenvironmental change. The analysis presented here addresses the question, what, if any, lithic assemblage traits at PP5-6 represent changing behavioral responses to the MIS 5-4-3 interglacial-glacial cycle? It statistically evaluates changes in 93 traits with no a priori assumptions about which traits may significantly associate with MIS. In contrast to other studies that claim that there is little relationship between broad-scale patterns of climate change and lithic technology, we identified the following characteristics that are associated with MIS 4: increased use of quartz, increased evidence for outcrop sources of quartzite and silcrete, increased evidence for earlier stages of reduction in silcrete, evidence for increased flaking efficiency in all raw material types, and changes in tool types and function for silcrete. Based on these results, we suggest that foragers responded to MIS 4 glacial environmental conditions at PP5-6 with increased population or group sizes, ‘place provisioning’, longer and/or more intense site occupations, and decreased residential mobility. Several other traits, including silcrete frequency, do not exhibit an association with MIS. Backed pieces, once they appear in the PP5-6 record during MIS 4, persist through MIS 3. Changing paleoenvironments explain some, but not all temporal technological variability at PP5-6.

Late Pleistocene and Holocene evidence from multiple rockshelters in north-central Tanzania suggests a regional pattern of changing technological behaviors through time. We use independent chronological evidence to test if the proposed technological patterns across space were also temporally equivalent. We applied AMS radiocarbon dating methods to the carbonate fraction of five ostrich eggshell fragments from Mehlman’s 1975-1976 excavations at Nasera rockshelter and compared our results to recent re-dating efforts of Mumba rockshelter. We document radiocarbon results >46 ka at Nasera in Level 5, indicating that associated and underlying technologies (including what Mehlman termed the ‘Nasera Industry’) are older than previously documented. Backed pieces first appear >46 ka at Nasera, which is in accordance with recent evidence from nearby rockshelters like Enkapune ya Muto, Panga ya Saidi, and potentially Kisese II. We also provide an age of 11,260-11,710 calBP for the ‘Silale Industry’ of Level 3B. Overall, the shifts in lithic technology previously considered to be shared between Mumba and Nasera rockshelters are not temporally aligned, emphasizing the possibility that intra-site variability was the norm throughout the Late Pleistocene in eastern Africa.