Female craniometrics support the ‘two-layer model’ of human dispersal in Eastern Eurasia

This study reports a cranio-morphometric analysis of female human remains from seven archaeological sites in China, Vietnam and Taiwan that date between 16,000 and 5300 BP. The aim of the analysis is to test the “two-layer” model of human dispersal in eastern Eurasia, using previously unanalysed female remains to balance the large sample of previously-analysed males. The resulting craniometric data indicate that the examined specimens all belong to the “first layer” of dispersal, and share a common ancestor with recent Australian and Papuan populations, and the ancient Jomon people of Japan. The analysed specimens pre-date the expansion of agricultural populations of East/Northeast Asian origin—that is, the “second layer” of human dispersal proposed by the model. As a result of this study, the two-layer model, which has hitherto rested on evidence only from male skeletons, is now strongly supported by female-derived data. Further comparisons reveal that the people of the first layer were closer in terms of their facial morphology to modern Africans and Sri Lankan Veddah than to modern Asians and Europeans, suggesting that the Late Pleistocene through Middle Holocene hunter-gatherers examined in this study were direct descendants of the anatomically modern humans who first migrated out of Africa through southern Eurasia.

www.nature.com/scientificreports/ but they are mainly focused on Neolithic and early Metal Age samples from Thailand, Vietnam and Laos, hence they are not directly relevant for the issues discussed here. In our previous study, analysis of cranio-morphometric patterns in eastern Eurasian and Sahul specimens (approximately 800 skeletons from late Paleolithic through Iron Age contexts) strongly supported the "two-layer model" of AMH dispersal in these regions 1 . According to this model, the "first layer", the original Late Pleistocene AMH colonizing population, shared a direct ancestry with present-day Indigenous Australians and Papuans. In Southeast Asia, this first layer is represented primarily by Hoabinhian hunter-gatherers whose skeletal remains date between 12,000 and 4000 BP, before the expansion of Neolithic farmers. In our previous study, we regarded the agricultural populations who cultivated rice (Oryza sativa; and some also cultivated millets) as the "second layer", who brought northeast Asian cranial features into Southeast Asia during the Neolithic period (from ca. 4000 BP).
Significantly, this two-layer model has hitherto been based only on male data sets. Male skulls were used for a variety of reasons, especially that morphological divergence is greater among male than female crania owing to greater male cranial size and robustness. Additionally, there are more male than female archaeological specimens available for study.
Accordingly, our new study examines female skeletons from seven Late Pleistocene through Middle Holocene archaeological sites in China, Vietnam and Taiwan ( Fig. 1: site location map). The oldest comes from Yahuai Cave (n = 1) in Guangxi Province, southern China. Nine more pre-farming individuals come from Huiyaotian (n = 6) and Liyupo (n = 3) in Guangxi Province, southern China. A further three come from Hang Cho (n = 1), Mai Da Dieu (n = 1), and Bau Du (n = 1) in Vietnam. One comes from Xiaoma Cave (n = 1) in southeastern Taiwan.
In the context of the two-layer model, these analyzed specimens are all presumed to have been members of the first layer of human dispersal. Our study compares the craniometric features of the aforementioned female specimens with those of contemporary populations in Africa, Eurasia, and Melanesia. Although many other female skeletons have been reported from Late Paleolithic, Neolithic and Iron Age sites across the region, the lack of standardization in the recorded measurements and the lack of published data make it difficult to include them. Instead, we use comparative data from modern populations measured according to our standard protocols.
The study sites. Table 1 presents details of the analysed remains and archaeological sites, including pertinent radiocarbon dates. Figure 2 exhibits the analysed cranial specimens unearthed from these focal sites. Sex was determined from pelvic and cranial morphology, and other post-cranial markers 9 . Sex determinations for new individuals reported in this study are described, otherwise details can be found in the cited publications.   (Fig. 3).
The main rockshelter and inner cave enclose a total area of more than 100 m 2 . Three seasons of excavation were conducted between 2015 and 2018 by an archaeological team led by co-author Xie from the Guangxi Institute of Cultural Relic Protection and Archaeology 10 . The excavation was divided into four areas (A, B, C, and D), with a total exposure of 50 m 2 . Total depths reached 7.5 m at the rock floor in Area A, but cultural deposits are up to 5 m thick, most dating from 44,000 through 16,000 BP by AMS 14C dating. A complete human skull with a partial mandible was discovered beneath layer 6 in Area B, and dated by three associated charcoal samples to 16,000 BP 10 ( Table 1). As the post-cranial skeleton was not exposed, sex  www.nature.com/scientificreports/ determination is based on the skull. The perpendicular and elevated frontal bone, small mastoid processes, and the smoothness of the occipital muscle attachment area are all female characteristics. Two hearths were found in the Pleistocene deposits at Yahuai. Tens of thousands of stone artifacts, some shell and bone tools, and an abundance of animal and plant remains (including wild rice phytoliths) were recovered. The lithic assemblage is dominated by small amorphous flakes and pebble tools of quartzite, sandstone, flint, tektite and quartz crystal 11 .
Other comparative hunter-gatherer sites. Huiyaotian Shell Midden, Guangxi Province, China. The well-preserved Huiyaotian Shell Midden is located on the first terrace of the Yongning River in Qingxiu District, Nanning City, Guangxi. The site was discovered in 1973 and first excavated by a joint archaeological team from the Guangxi Institute of Cultural Relics and Archaeology (currently the Guangxi Institute of Cultural Relic Protection and Archaeology) and the Nanning City Museum in 2006. Co-authors Li, Matsumura, Hung and L. C. Nguyen then conducted further research in 2012 12 . The thickness of the cultural component is about 60-120 cm, in a matrix primarily composed of river shells. Three samples, including two charred Canarium sp. seed fragments and one human tooth, produced AMS C14 dates of 9000-8300 BP ( Table 1). The assemblage belongs to the Dingsishan Cultural Phase in the Yongning River Basin, which pre-dated the arrival of rice-farming economies from the Middle-Lower Yangtze Basin 3,5 . In total, 64 inhumations were recovered, including flexed, squatting and some dismembered burials, all without grave goods. The artifacts recovered in association with the human remains included polished stone axes and adzes; a bone assemblage of adzes, arrowheads, awls, needles and fish-hooks; shell tools; and sherds of cord-marked pottery.  www.nature.com/scientificreports/ The excavation area in 2009 included three trenches totalling 26 m 2 . The deposit contained shells and artifacts, including polished stone axes and adzes, grindstones, small bone awls, and shell shovel-like artifacts. AMS C14 dates on charred Canarium sp. seeds, charcoal, and human bones and a tooth give a chronology of c. 8000-6700 BP (Table 1). In total, 40 flexed skeletons were unearthed, many deliberately protected under large stones 12 .
Hang Cho Cave, Hoa Binh Province, Vietnam. Hang Cho is a limestone cave site located near Cao Ram Village, Luong Son District, Hoa Binh Province, northern Vietnam. French archaeologist Madeleine Colani led the primary excavation in l926 and 1932, uncovering Hoabinhian stone tools 13 . Further investigation occurred by the Vietnamese Institute of Archaeology in l997-1998. In 2003, a research team from Hanoi National University and Seoul National University obtained freshwater shell AMS C14 dates of c. 19,500 and c. 8400 BP 14 .
In 2004, another multinational excavation project including Matsumura and L. C. Nguyen uncovered a flexed human skeleton (04HCH3M1) at the cave entrance 15 . Its pelvis has an obtuse-angled greater sciatic notch that suggests a female. Direct AMS radiocarbon analysis of a single tooth dates the skeleton to approximately 11,000 BP (Table 1). A total of 1523 Hoabinhian lithics were found in stratigraphic association with the skeleton, including a few edge-ground axes as well as pebble and flake tools.  [16][17][18] . The site contains a long cultural sequence with C14 dates ranging between 19,700 and 3500 BP. Twenty six human burials have been recovered from the site, among them a young adult (84MDD1) found in well-preserved condition 0.7 m below the surface. The pelvis of this individual has an obtuse-angled greater sciatic notch, suggesting a female. Indirect C14 dating on charcoal indicates an age of c. 8000 BP for this individual, within a Hoabinhian archaeological context (Table 1).  (Table 1).

Bau Du Shell
Bau Du produced pebble and flake tools, but no polished tools and no pottery. Several flexed human skeletons were found sealed in pits beneath the shell midden. This study examines the nearly complete skull of one of them (14BDHIL5M4) discovered in 2014 20 . The sex assignment as female is based on its obtuse-angled greater sciatic notch.
Xiaoma Cave, Taidong, Taiwan. Xiaoma is a limestone cave located in Chenggong Town, Taidong County, southeast Taiwan. The site was excavated in 1988-1990 by Huang and the National Taiwan University archeological team [21][22][23] . The excavation uncovered a flake tool assemblage, probably related to the late Changbin (Preceramic) cultural phase of east coastal Taiwan, and a burial pit dated on marine shell to c. 6200-5700 BP ( Table 1). The Xiaoma human burial is the only known pre-Neolithic burial from the main island of Taiwan, and it was placed in a squatting position. Following its excavation in 1988, it was stored, without cleaning or reconstruction until this study, in the Anthropology Museum of National Taiwan University.
The short stature, short limbs, and small cranial size of the Xiaoma individual indicate a close affinity to Negrito (Ayta) groups in the Philippines. Although the Austronesian-language oral traditions of Formosan populations contain references to the existence of small-statured and dark-skinned people on the island at some time in the past 24 , this reference never has been confirmed by skeletal data, until now. Therefore, Hung, Matsumura, and L. C. Nguyen initiated a joint project on this skeleton in 2016, resulting in the reconstructed cranium presented here. Although the innominate bone of this individual is too damaged for sex determination, it is assumed to be female because of its gracile cranium, small mastoid processes, smooth occipital muscle attachment area, perpendicular and elevated frontal bone, and the smooth contour of its mandibular base.

Results
Craniometric affinities within eastern Eurasia and Melanesia. Our study commenced with an analysis of the female skulls just listed, together with others from mid-Holocene burials in Jomon Japan, in order to reveal cranial affinities within this population assemblage. Figure 4 depicts a tree diagram based on the Neighbor Net analysis of Q-mode correlation coefficients using 13 cranio-morphometric datasets (Tables S1 and S2). The diagram exhibits two major clusters. The cluster on the upper left comprises Northeast and Eastern Asians (NEA) as well as present-day Southeast Asians (SEA). The cluster on the lower right comprises Indigenous Australians and Melanesians (MEL, including Papuans).
Craniometric affinities across Eurasia, Melanesia and Africa. As Fig. 5 shows, analysis using a much larger data-set encompassing western Eurasian and African populations (Tables S3 and S4)  The main observation to be drawn from this distribution is that the Australo-Papuan, South Asian Veddah, Nicobar Island and ancient Jomon samples, plus the seven ancient samples analysed in this study, have a closer affinity to African samples than they do to any other Eurasian samples included in this study.
Principal component analysis. The PCA results using the mean data-sets of 13 standardized cranial measurements taken from 67 modern population samples are given in Table S5, with their eigenvectors, eigenvalues and loadings. Table S6 provides the PCA scores for all comparative population samples, including the focal prehistoric series. The loading values are of the first five components, the eigenvalues of which are greater than 1, and the cumulative contribution is 85.64%. Figure 6 represents the loading values of the 13 cranial measurements for the first four principal components whose contribution rate is greater than 10%. The first component (PC 1) reflects the overall cranial size because all measurements almost equally indicate positive loading values.
In the second PCA component (PC 2), the cranial breadth is fairly positive, while the length of the cranial vault and the breadth of the facial segments are negative, thereby suggesting that the cranium with lower PC scores will be characterized by a broader face and longer head. The third component (PC 3) exhibits relatively high positive loadings for the facial and nasal height against faintly negative loadings of breadth of overall segments. A skull with a lower negative score in PC 3 will possess a broad lower face. Meanwhile, the loadings of PC 4 tend to be positive for bimaxillary and nasal breadth and negative for cranial vault length and orbital size. Accordingly, specimens with high scores will be characterized by a wide nasal opening against orbital size. Figure 7 presents comparative samples, with the most negative scores in PC 1 calculated using the 13 cranial measurements. The cranial sizes of the Santa Cruz Melanesians, Veddah in Sri Lanka, Negrito in the Philippines, and San in South Africa are small, in accordance with their small body sizes-known features of these populations. Notably, the small size of the Xiaoma crania from Taiwan is comparable with these samples. Figure 8 plots the scores of PC 2 and PC 3 in a two-dimensional scheme. Northeast/Southeast Asians and Europeans are scattered on the right side, while Africans and Australo-Papuans-possessing a broader face and longer head-are clustered on the left side. With the exception of Xiaoma and Yahuai, the five focal specimens are In terms of un-schematized PC 4, as given in Table S6, the seven focal series indicate high positive scores, especially for Xiaoma and Bau Du, characterizing a wide nasal opening against orbital size.

Discussion
As mentioned, in a previous study on the dispersal process of AMH in eastern Eurasia, we proposed the two-layer model based on cranial data-sets of male specimens 1 . We proposed that the Late Pleistocene AMH colonizers (first layer) in southern China and Southeast Asia-akin to the ancestors of current Indigenous Australo-Papuans-were encountered during the Holocene by a second layer of migrants from northern Eurasia who possessed morphological characteristics adapted to a cold climate. The dispersal of this second layer of human population involved large scale demic expansion of agriculturists from the Neolithic period onwards.
This two-layer model of population history has been strongly supported by recent genome-wide ancient DNA analyses in East/Southeast Asia [25][26][27][28] . Indeed, these genomic studies suggest that the first layer, which equated archaeologically with the Hoabinhian in Mainland Southeast Asia, shared genetic resources with the ancient Jomon people of Japan 26 , as documented independently by the morphometric cranial affinities between these two populations that we have discussed above.
Against the two-layer model, the "regional continuity" Sundadont/Sinodont model proposes that modern Northeast Asians derive from Southeast Asian migrants, the reverse of what is suggested by us. The regional continuity model is not supported by our previous dental 29 or male craniometric data 1 , or by our current study based on female craniometric data. Additionally, it is not corroborated by recent genetic studies 30 .
Our study shows that pre-Neolithic female hunter-gatherers from our sample of seven sites in southern China, Taiwan and Vietnam, as well as the Holocene Jomon population of Japan, all belonged to the first layer within our two-layer model. Our findings demonstrate a close cranial affinity between the analysed specimens and recent Indigenous Australians and Papuans. In terms of the Huiyaotian and Liyupo samples, the clustering pattern generated by female craniometric data is consistent with that based on male data in previous studies 1,12 , further underscoring the significance of the cranial affinity.
A comparison of the specimens that we have analysed with a larger population that extends as far as western Eurasia and Africa produces a more comprehensive interpretation. The pre-Neolithic skulls and recent www.nature.com/scientificreports/ Australo-Papuan samples exhibit a closer resemblance to African samples than do present-day Eurasian samples. In addition to the morphometric distances produced by the Q-mode correlations, PCA analysis has also revealed the cranial affinities that exist between our sample of pre-Neolithic skulls and recent Australo-Papuans and Africans. Cranial morphological affinities between early Africans who lived at the time of dispersal into Eurasia and present-day Africans remain obscure, because informative African samples from the period around 60-50 ky are lacking. Available reports on AMH cranial remains from this period 31-34 describe archaic or hybrid archaic/ modern features, but those remains are fragmentary. Nevertheless, the close affinity between modern Australo-Papuans and modern Africans supports our hypothesis that the first AMH populations to leave Africa successfully were similar in cranial morphology. In this respect, present-day Eurasian people possess a different cranial shape generally attributed to a dry and/or cold climatic adaptation.
A number of different "Out of Africa" models based on skeletal and genomic data have been proposed to explain AMH dispersal. The simplest presents a single dispersal from Africa into Eurasia along migration routes that avoided geographic barriers like the Himalayas [35][36][37][38] . Another model proposes multiple waves of migration, with the first dispersal from Africa occurring along the Indian Ocean Rim into Southeast Asia and Australia, followed by a subsequent dispersal to northern Eurasia [38][39][40][41][42][43][44] . An initial AMH migration that followed the coastal rim of the Indian sub-continent and continued through Southeast Asia would link well with the later emergence of the archaeologically-defined Hoabinhian stone tool complex in this region.
There is one other point to be emphasized about our findings concerning the Xiaoma cranium from Taiwan. PCA analysis revealed that, in respect to overall cranial size, Xiaoma is very small (PC 1), and comparable to Negrito crania from the Philippines, San from South Africa, and Veddah from Sri Lanka. These people are all known for their short stature and small body size. The Philippine Negrito sample is assumed to belong to our first layer because of its non-metric dental traits, which are highly heritable 29 . Modern Negrito samples included in Moreover, the small cranium of the Xiaoma individual from Taiwan, together with its apparent short stature from the surviving long bones (a precise estimate of stature is not provided due to incompleteness), supports many Formosan Austronesian oral traditions about the former existence of small and dark-skinned people in Taiwan 24 . Our study not only solves a long-standing mystery in Taiwan society, but also it provides significant www.nature.com/scientificreports/ insight into the population structure of prehistoric Taiwan, an island that is regarded by linguists as the homeland for the Austronesian language family and many of its speakers 45 . Our craniometric analysis is the first to be based on female skeletal data from sub-tropical and tropical southern China, Taiwan, and Southeast Asia. The prevailing hot and humid climate offers poor preservation conditions for ancient human burial remains, such that any discoveries can be regarded as precious for anthropological study. Our research indicates the great potential of heretofore overlooked female skeletons for craniometric analysis. The numerous female cranial remains from Neolithic, Bronze, and Iron Age contexts in China and Southeast Asia should also be included in future systematic studies.

Materials and methods
This study examines female skulls discovered in seven hunter-gatherer sites in southern China, Vietnam, and Taiwan, as listed in Table 1. These female skulls are compared with baseline data observations from ancient Jomon hunter-gatherers (c. 5000-2300 BP) in Japan 46 , and further compared with present-day population samples. The results allow assignment of the samples to their respective layer in the two-layer model. Table 2 presents the cranial measurements used in our study, based on Martin's definitions 47 .
The measurements of the comparative cranial samples from modern Eurasian, Melanesian, and African populations, as well as the Jomon sample, were taken from the worldwide database of cranial and dental characteristics compiled by co-author T. Hanihara before 2004. Details of the primary sources used by Hanihara and where samples are stored can be found in his previous publications 38,48,49 . When computing Q-mode correlation coefficients between population samples, the datasets were standardized using the grand means of comparative population samples and standard deviation data from the Jomon sample (Tables S1 and S3). www.nature.com/scientificreports/ In order to facilitate the interpretation of phenotypic affinities between samples, we used the Neighbor Net Split method 50 to the distance (1 − r) matrix of the Q-mode correlation coefficients (r) using SplitsTree, Version 4.05 45 .
Finally, in order to interpret cranial morphometric factors, we conducted principal component analysis (PCA) using the mean value datasets of 13 cranial measurements on all comparative population samples.

Ethics declarations and research permits.
The research permit for each specimen was issued from the associated repository authority. Yahuai (n = 1) and Liyupo (n = 3; M22, M23, M28) are kept in the archaeological storage of Guangxi Institute of Cultural Relic Protection and Archaeology, Nanning, Guangxi, China; Huiyaotian (n = 6; M7, M19, M20, M32, M39, M45-2) are in the Nanning Museum, Nanning, Guangxi, China; Hang Cho (n = 1; 04HCH3-M1) and Mai Da Dieu (n = 1; 84MDD-M1) are in the Institute of Archaeology, Vietnam Academy of Social Science, Hanoi, Vietnam; Bau Du (n = 1; 14BD1-M4) is in the Quang Nam Provincial Museum, Quang Nam, Vietnam; Xiaoma Cave (n = 1; C5) is in the NTU Museum of Anthropology, National Taiwan University, Taipei, Taiwan. The individual (G. Xie, the second author of this study) in the Fig. 3 has given written informed consent to publish these case details.

Data availability
All of the raw data used in this study are available in Table 2 and the "Supplementary Files" (Tables S1-S6), available online with this manuscript.