Main

Genetic, anthropological and archaeological studies support an African origin of Homo sapiens1, but the evolutionary process is debated based on fossils, archaeology and genetics2,3,4,5,6,7, with Africa harbouring the greatest human genetic diversity8,9, and southern and central African hunter-gatherer groups displaying some of the deepest diverging Homo sapiens lineages7,10,11,12. Population stratification between southern Africa (the region south of the Zambezi River) and the rest of Africa probably existed for at least 300 thousand years (kyr)4,5,7,13, perhaps up to a million years6. Such deep stratification may result from admixture with an unknown archaic African group predating the divergence of Homo sapiens from Neandertals and Denisovans1,5, and/or from isolation from other groups.

All investigated modern-day central and southern African Indigenous groups show substantial mixing with western and eastern Africans4,7,10,14,15,16,17, following the large-scale migrations starting 5 thousand years ago (ka) that veil more ancient events1, making it difficult to assess deep human evolutionary history using genomic data from modern-day people. By investigating variation among individuals living before recent large-scale population movements/admixture, palaeogenomic approaches overcome this limitation. Although they are restricted to few individuals and/or single archaeological sites with limited genomic data, these studies show longstanding stratification of ancient ancestries in eastern18,19,20, western21, northern22,23 and southern Africa4,17,19.

We sequenced the genomes of 28 ancient individuals from south of the Limpopo River (South Africa), all dating to the Holocene epoch, with Later Stone Age and Iron Age archaeological affiliations (Fig. 1 and Supplementary Information 1.1). Sampling was geographically and temporally broad (Fig. 1c), with remains recovered from sites across the southern and central parts of South Africa (Fig. 1b). From Matjes River on the southern coast, we sequenced the genomes of individuals spanning around 8 kyr (10,200–2,330 cal. bp; Supplementary Information 1.2).

Fig. 1: Archaeological sites and dates of sampled and comparative individuals.

a, All ancient African individuals from >1 ka with complete sequenced genomes (>7× coverage). The seven complete ancient genomes from southern Africa (Matjes River 1, 3 and 11, Great Brak River Cave, Cape St Francis, Springbokvlakte and Ballito Bay A) are studied in detail. The symbol shape indicates archaeological context for southern Africa (see the legend in b). The Zambezi and Limpopo rivers are highlighted in blue. b, Ancient African individuals with genomic data south of the 14th latitude. Archaeological context is given by the shape of the symbol. LSA, Later Stone Age. c, The distribution of sampled and comparative individuals across dates. Maps were created with Natural Earth in R, using vector maps from the package rnaturalearthdata.

The sequenced ancient individuals (Fig. 1b) range in coverage from 25.1- to 0.002-fold, with six individuals (Fig. 1a) having over 7.2-fold coverage (Extended Data Table 1). Radiocarbon dates and dietary isotopes were generated for most individuals (Extended Data Table 1, Supplementary Information 2.2 and 3.1, Supplementary Fig. 1 and Supplementary Data 1 and 2). Chronologically, the data cover five archaeological phases—Oakhurst (n = 1), Wilton (n = 4), Final Later Stone Age (n = 3), Ceramic Final Later Stone Age (n = 18)24 (Supplementary Information 1.1) and Iron Age (n = 2), spanning between 10.2 ka and a few hundred years ago. To assess the relationship between the ancient southern Africans (Supplementary Data 1 and 5) and other groups, we co-analysed our data with relevant ancient and modern-day Africans, as well as relevant ancient and/or modern-day Europeans, Asians, Americans and Oceanians (hereafter collectively referred to as non-Africans; Supplementary Data 7 and 8).

Most ancient southern Africans (25 out of 28) carried haplotypes belonging to the mitochondrial L0d haplogroup (Extended Data Table 1), which is common among modern-day Khoe-San individuals25 (Supplementary Information 1.3). The oldest individual (Matjes River 6) and four other individuals (Great Brak River Cave, Cape St Francis and two from Ballito Bay4) carried a Y chromosome with haplogroup A1b1b2a, which is also common among modern-day Khoe-San individuals26. Two individuals who lived ≤500 years ago displayed a Y chromosome with haplogroup E1b1b1, which is common among eastern Africans and modern-day Khoe-San people26, consistent with recent gene flow into southern Africa with eastern African pastoralists4,19. Variants for high skin pigmentation, brown eyes and non-lactase-persistence were fixed among the seven complete ancient southern African genomes (Fig. 1a and Supplementary Data 6). None of these carried the Duffy-null variant that is protective against Malaria (rs2814778) or the G-variant at rs73885319 in the APOL1 gene that is protective against sleeping sickness (Supplementary Data 6), even though these variants were present in the region around 500 years ago among individuals with western African ancestry4.

Unique ancient southern African ancestry

Model-free (principal coordinate analysis, PCoA) and model-based approaches (Supplementary Information 2) demonstrate (Fig. 2a and Extended Data Fig. 1) that the first two dimensions (explaining the most genetic variation) form a V-shaped pattern of non-Africans at one end of the distribution, through a gradient of eastern Africans (modern-day and ancient) towards western and central Africans in a second end of the distribution. Indigenous southern Africans form the third end7,8,27. Notably, many of the ancient southern Africans (including all individuals between 10.2 ka and 1.4 ka), fall outside the range of genetic variation among modern-day individuals (including Khoe-San groups), and form an extreme end of human genetic variation (at axis 1 and axis 2). This pattern is veiled if the ancient genomes are projected on top of axes of variation built from modern-day genomes and/or ascertained/captured single-nucleotide polymorphisms (SNPs)17 (Supplementary Information 2.7 and 3.8–9 and Supplementary Figs. 2, 3 and 9). The ancient western African individuals from Shum Laka dating to 7.9–3.1 ka (ref. 21) are located close to modern-day central African rainforest foragers (for example, Biaka, Mbuti) and western Africans. Ancient eastern18 and east-central African individuals20,28 are distributed among modern-day individuals from eastern, western and central Africa, with ancient eastern African Neolithic pastoralists28 clustering with the modern-day eastern African Amhara people. Ancient individuals from southeastern Africa dated to between 8,000 and a few hundred years ago19,20 (Malawi, Zambia) cluster in-between eastern Africans (ancient and modern day) and ancient southern Africans.

Fig. 2: Characterization of the genomic variation in modern-day and ancient Africa.

a, PCoA of the 28 ancient southern Africans in this study (pink) and all comparative ancient individuals with genetic information across sub-Saharan Africa (modern-day individuals from southern, southeastern, eastern (E), central (C) and western (W) Africa; individual labels are shown in Extended Data Fig. 1). b, Gene flow (estimated by two different f4 tests) from an eastern or western African source into the ancient southern Africans, plotted against years bp and latitude south of the Equator. The vertical dotted line marks 1,300 cal. bp, when admixed individuals first appear. c, Estimated ancestry, assuming five ancestry components. The results assuming between two and ten ancestry components are provided in Extended Data Fig. 2 and Supplementary Fig. 6. d, Pairwise genetic differences visualized using a hierarchical clustering algorithm (UPGMA) among all pairs of ancient (>1 ka) high-quality genomes (>10×) from Africa and a selection of comparative high-quality ancient genomes from elsewhere (including three Neandertals and a Denisovan individual). bp, base pair. e, Genetic diversity (heterozygosity) for ancient high-quality (>10×) genomes.

Assuming two ancestry components (Supplementary Information 2.5 and Supplementary Data 7 and 8), one was anchored among the ancient southern Africans (whose genome displays 100% of this ancestry component) and the other among non-Africans (ancient and modern-day Europeans). All of the other individuals (including modern-day and ancient western, eastern and central Africans) were distributed with varying fractions of these ancestry components (Extended Data Fig. 2 and Supplementary Fig. 6), reiterating the unique position of ancient southern Africans. Assuming 3–5 ancestry components (Fig. 2c and Extended Data Fig. 2), the genomes of individuals from western Africa, central Africa and eastern Africa were largely assigned to these ancestry components. Assuming larger numbers of ancestry components typically separated out individual populations (Extended Data Fig. 2 and Supplementary Fig. 6), representing finer-grained population stratification. Assuming five ancestry components (Fig. 2c), the genomes of all ancient southern African individuals between 10,200 and 1,400 cal. bp consist completely of the ‘ancient southern African ancestry component’ (pink), with no indications of admixture (Fig. 3 and Extended Data Fig. 6). For these individuals, there is no indication of temporal genetic stratification (r2 = 0.02, P = 0.06), and low levels of spatial stratification (r2 = 0.07, P = 0.001, Supplementary Information 3.8 and Supplementary Figs. 7 and 8) despite spanning around 9 kyr and a vast landscape (Fig. 1b). Using the oldest high-coverage individual (Matjes River 1) as the anchor individual in a test based on the fraction of shared derived variants to measure population continuity (Supplementary Information 2.16), we find a pattern of population continuity in southern Africans spanning 7 kyr at least (Extended Data Fig. 3), although all individuals older than around 3 ka come from Matjes River.

Fig. 3: Graphical representation of the inferred human population history of southern Africa.

The summary of observations from the ancient genomes in this study that are shown in Figs. 1b,c and 2a–c. Individual symbols correspond to the symbols in Fig. 1b. R., river.

Average population divergences between individuals representing the ancient southern African group (7 individuals with >7.2-fold genome coverage) to any other individual (ancient and modern-day western, eastern, central, northern Africans and non-Africans) were estimated to around 310–240 ka using a two-by-two site-frequency spectra approach (Supplementary Information 2.17 and Supplementary Data 15–31). Although the exact calibration of chronological population divergence-time estimates depends on model assumptions, mutation-rate assumptions and generation time5, these estimates recapitulate findings in which the divergence between ancient southern Africans and all other groups captures the deepest population split-time at around 300 ka (refs. 4,5,7,11,13,16) (see also Supplementary Fig. 11 for a comparison to modern-day Khoe-San individuals). This approximately 300 ka population-divergence-time estimate is not caused by a deeper partial archaic admixture event per se, but it does not negate such an event either6,29 (Supplementary Information 3.11 and Supplementary Figs. 12 and 13).

We further found that the genetic affinity between ancient southern Africans and eastern Africans (ancient and modern day) is similar to the genetic affinity between ancient southern Africans and western Africans (ancient and modern day; for example, f4[Denisova, Matjes River 1, Mota, Shum Laka] ≈ 0; Supplementary Fig. 5), suggesting that detectable gene flow from either of these groups was unlikely since around 150–200 ka (the time that marks the divergence between western and eastern Africans; (Supplementary Information 3.5–3.6 and Supplementary Data 25).

Southern ancestry further north

However, the southern African genetic component was present further north by the mid-Holocene19,20,30. Individuals at 11–12° S (current-day Malawi, Zambia) display a mixture of eastern and southern African ancestry by 8 ka (refs. 19,20), but no individual with a majority of southern African ancestry has been found in this area. Modern-day Juǀ’hoansi northern San traditionally living around 18–25° S (Botswana, Angola, Namibia), also show eastern African admixture (11%) that may have occurred at the onset of the Holocene13. South of 15° S, we detect admixture from eastern and western African sources only from around 1.3 ka (Fig. 2b and Extended Data Fig. 6), but note that few humans dated to >1.3 ka have been palaeogenetically investigated from 10–35° S.

Gene flow into the south since 1.4 ka

Among the three individuals who lived between 1.3 and 0.7 ka, two (Koffiefontein, Kasteelberg) show a distinct ancestry component matching eastern African Neolithic pastoralists (Fig. 2c (purple and green)), demonstrating that this gene flow extended over a large area. Among the southern Africans living between 0.6 and 0.1 ka, we note: (1) two individuals (Bushveld, Vaalbank) displaying only the ancient southern African ancestry component (Fig. 2c (pink)); (2) eight individuals with a majority of the ancient southern African ancestry component (pink), and with some level of an eastern African (purple and green) or European (green) ancestry; and (3) six individuals who display a substantial western African ancestry component (yellow), with minor fractions from the ancient southern African ancestry (pink), eastern Africa (purple) and sometimes with a European ancestry component (Plovers Lake 1, Tobias Cave 1).

Our findings therefore contrast with linguistic, archaeological and some early genetic studies pointing to a shared ancestry or long-term interaction between eastern, western and southern Africa. For example, it was suggested that the contemporary southern African Khoe-San people are the descendants of a once-widespread population that extended across much of southern, eastern and northeastern Africa31. Instead, there was deep population stratification (Figs. 2 and 3), with gene flow reaching south of 15° S only by 1.3 ka. The proposed linguistic connection between eastern African Hadza people and southern African click languages has been refuted, and the connection between eastern African Sandawe people and southern African click languages is now seen as resulting from the introduction of pastoralism to southern Africa after 2 ka (ref. 32).

Partial continuity into Khoe-San

Modern-day Khoe-San groups show ample recent admixture with western African, eastern African and non-African groups. In 25 Khoe-San genomes selected for ‘least admixture’7, we estimate an average ancient southern African ancestry of around 79%. The last unadmixed individuals (in our sequenced data) seemingly disappeared a few centuries ago (Fig. 2c). Modern-day Juǀ’hoansi (around 11% eastern African admixture) and Karretjie People (around 17% mainly western and eastern African admixture) show the greatest genetic similarity to the ancient southern Africans (Fig. 2c, Extended Data Fig. 2 and Supplementary Fig. 6). However, the ancient southern Africans were substantially genetically differentiated from Juǀ’hoansi individuals (Wright’s fixation index* F*ST = 0.055; similar to modern-day Finnish versus Chinese individuals; Supplementary Data 9). This differentiation is not solely due to admixture from eastern Africans—it also reflects the deep stratification of northern and southern San groups4,7,27 (Extended Data Fig. 2). Genetic differentiation was also distinct between ancient southern Africans and Karretjie People (FST = 0.041; similar to modern-day Indian versus Chinese individuals; Supplementary Data 9), consistent with partial continuity and recent admixture from western and eastern African groups.

Long-term large population size

For the ancient African genomes (over tenfold coverage), as well as four pre-Neolithic Eurasians, three Neandertals and one Denisovan individual (Supplementary Information 2.5), pairwise genetic differences for the full genomes (π; Extended Data Fig. 5) showed around 1.41 differences per 1,000 bp (π = 1.41 × 10−3) between Homo sapiens and archaic humans (Fig. 2d). Pairwise differences between ancient southern Africans and other ancient Africans and pre-Neolithic Eurasians were slightly lower (π = 1.02 × 10−3) than between Neandertals and Denisovans (π = 1.14 × 10−3). Pairwise differences within the group of ancient southern Africans (π = 0.82 × 10−3) was slightly greater than between ancient western and ancient eastern Africans (π = 0.79 × 10−3). Heterozygosity (HO) for ancient southern Africans (mean across genomes; HO = 0.80 × 10−3) was similar to other ancient Africans, only surpassed by an ancient western African individual (HO = 0.93 × 10−3), indicating a large Holocene population size in southern Africa (Fig. 2e). A multiple sequentially coalescent approach (Supplementary Information 2.11) shows that the effective population size (Ne) was large for several hundred thousand years, up to Ne ≈ 30,000 around 200 ka (Supplementary Fig. 10), similar to other African groups7. The large Ne at ≥300 ka for all humans was potentially caused by population subdivision29. We note a decline in Ne for ancient southern Africans from around 100–50 ka, to Ne ≈ 10,000 by the Last Glacial Maximum (20 ka), similar to non-African groups and the ancient northern Africans22.

Runs of homozygosity (ROH, where greater numbers and total length of ROH segments indicate a smaller population size; Extended Data Fig. 4) show that the ancient southern Africans were at the upper tail of the distribution of modern-day Africans, but less extreme than most non-Africans—a pattern attributed to the out-of-Africa bottleneck. This indicates a smaller population size (relative to, for example, western African groups) in the relatively recent history of each individual, but still larger compared with non-Africans and ancient northern Africans (Extended Data Fig. 4). Most ancient southern Africans are shifted towards greater total segment ROH length without affecting the total number of ROH segments (shifted off the diagonal line in Extended Data Fig. 4), in particular the Great Brak River (2,355–2,310 cal. bp) and the Matjes River 1 (7,845–7,690 cal. bp) individuals. This pattern indicates a smaller recent ancestral population size, possibly with elements of inbreeding, indicating isolation and fragmentation among ancient southern Africans during the Holocene (see Supplementary Information 4.1 and Supplementary Fig. 4 for diet variability). Ancient southern Africans south of the Limpopo River therefore consisted of a large, stable population for many millennia, with a modest decline since around 50 ka, and a possible fragmentation and further decline during the Holocene.

Southern Africa as a long-term refugium

Cultural contact between southern African foragers and incoming farmers is detected archaeologically from around 2 ka (refs. 24,33). The palaeoanthropological record shows a similar transition pattern in the late Holocene, but also a distinct difference in fossils (for example, the Hofmeyr cranium) compared to before the Last Glacial Maximum34 (Supplementary Information 4.1). Genomic data are compatible with southern Africa serving as a geographical refugium to a large human population for several hundred thousand years, affected only by gene inflow in the last millennium4,19 (Fig. 2). However, gene outflow from southernmost Africa probably happened in pulses during favourable climatic conditions35, potentially already around 70 ka (ref. 30) (Fig. 3). This resembles an ‘isolation by fragmentation’ model in which human groups were separated in different African refugia for extended periods35, as opposed to an ‘isolation by distance’ model with maintained gene flow across the continent. It appears that during favourable conditions, when the isolation-by-fragmentation ceases, gene flow is one-directional into intermediate areas that are depopulated for extended periods of less favourable conditions.

Alternatively, the southern African group inhabited the region up to 11–12° S before the Holocene, followed by an expansion of eastern African groups during Holocene, first to south-eastern Africa (as observed in the 8 ka mixed-ancestry genomes of Malawi)19 and reaching southernmost Africa in the last millennia. Either way, the genomic data of ancient southern Africans with its distinctiveness and lack of gene inflow before around 1.3 ka suggest that southern Africa represents a long-standing human refugium, with potential gene-outflow pulses. Although direct comparisons of complete genomes do not rule out complex demographic histories further back in time, nor low levels of archaic admixture, the deep human history in Africa can be represented by deep stratification—including long-term isolation—between southernmost Africa on the one hand and western, central and eastern Africa and the rest of the world on the other (Fig. 3).

Homo sapiens-specific genetic variants

Early Homo sapiens fossils with archaic and modern features dating to around 300–190 ka were found in southern36,37, eastern38,39 and northern Africa2. Some of the oldest Homo sapiens fossils with modern features come from Border Cave, South Africa, at around 171–152 ka (ref. 40), and Omo Kibish in Ethiopia at around 230 ka (ref. 41). Thereafter, all human remains in southern Africa are modern24, with some of the earliest archaeological evidence of modern human behaviour/thinking from at least 100 ka (ref. 42).

As behavioural and cognitive traits are largely heritable, southern Africa’s deep Homo sapiens genomic record could aid in disentangling the ‘sapient paradox’, whereby anatomical modernity purportedly predates modern behaviour43. However, this may not be straightforward, because some genes governing cognition and anatomy evolved rapidly in the early human lineage12, and genetic variants governing rapid neuron developments thought to be fixed in humans44 are highly variable in some populations45. Yet, our understanding of human cognitive evolution increases with cataloguing genetic variants associated with cognitive traits[46](https://www.nature.com/articles/s41586-025-09811-4#ref-CR46 “Kuhlwilm, M. & Boeckx, C. A catalog of single nucleotide changes distinguishing