Peopling the Americas: Not “Out of Japan”

ABSTRACT A widely accepted model for the peopling of the Americas postulates a source population in the Northeast Asian maritime region, which includes northern Japan. The model is based on similarities in stone artifacts (stemmed points) found in North American sites dating as early as 15,000 years ago and those of comparable age in Japan and neighboring regions of Northeast Asia. Here we show, on the basis of data and analyses in biological anthropology, that the people who made stemmed points in northern Japan (labeled “Incipient Jomon” in the archaeological literature) represent an unlikely source population for the indigenous peoples of the Western Hemisphere.


Out of Japan?
Despite a recent string of spectacular archaeological finds and the transformational impact of human paleo-genomics, many details regarding the peopling of the Americas remain unresolved. Recent discoveries have added considerable complexity to proposed models. These include terminal Pleistocene burials in central Alaska (Potter et al. 2014), 13,000-year-old footprints on the Pacific coast of North America (McLaren et al. 2018), a 14,500-year-old mastodon kill in northern Florida (Halligan et al. 2016), 15,000-year-old stemmed points in the Northwest region of the USA (Davis et al. 2019;Jenkins et al. 2012), and traces of an equally ancient coastal economy in Peru (Dillehay et al. 2017).
The location of the Eurasian source population(s) of the First Peoples of the Western Hemisphere, and the time at which the Native American founder group(s) diverged from that population (or populations), remain points of contention, as does the number and timing of migration events (e.g., Moreno-Mayar et al. 2018;Ning et al. 2020;Reich 2018;Sikora et al. 2019;Yu et al. 2020). And if most anthropologists accept Beringia as the general region through which the founder groups moved into the Americas, both archaeologists and geneticists debate interior versus coastal routes (e.g., Fagundes et al. 2008;Graf and Buvit 2017;Perego et al. 2009;Potter et al. 2018;Scheib et al. 2018).
One of the more widely accepted models of Native American origins postulates a source population in the Northeast Asian maritime region and a coastal migration around the North Pacific Rim roughly 16,000-15,000 calendar years ago (cal yr BP) (e.g., Davis et al. 2019;Erlandson and Braje 2011). Like most models for the peopling of the Americas, it is based primarily on the analysis of archaeological data (e.g., Meltzer 2009;Potter et al. 2017;Stanford and Bradley 2012;West 1996). The original concept was proposed by N. N. Dikov (1979, 31-53), who discovered stemmed points on Kamchatka (i.e., North Pacific Rim) in the 1960s and 1970s and noted similarities to Western Stemmed points in mid-latitude North America. In recent years, the model has been bolstered by some of the spectacular discoveries mentioned above, as well as supporting data for early deglaciation of the Pacific Northwest coast and corresponding delay in the availability of an interior route (Darvill et al. 2018;Lesnek et al. 2018;Pedersen et al. 2016).
The model is grounded in similarities of stemmed points found in Japan and other parts of the Northeast Asian maritime area, including Korea and the Russian Far East, classified archaeologically as "Incipient Jomon" and dating to roughly 16,000-14,000 cal yr BP (e.g., Nagai 2007;Ono et al. 2002) with those of comparable age in interior Western North America (e.g., Cooper's Ferry, Idaho) and somewhat younger along the Pacific coast of North America (e.g., Arlington Springs on the Channel Islands, California), 1 as well as the Pacific coastal zone of South America (e.g., Quebrada Jaguay in Peru) (Davis et al. 2019, 895;Erlandson 2013, 128-129;Erlandson and Braje 2011, 33-35;Sutton 2017Sutton , 2018 (Figure 1). 2 A "dual structure model" for the peopling of Japan is hypothesized by anthropologists, where the contemporary populations of Japanincluding the mainland Japanese, the northern Ainu, and the southern Ryukyuanare composed of two historical components: a very old hunter-gatherer population associated with Jomon pottery, spanning 16,000-2300 cal yr BP with even deeper roots in the early Upper Paleolithic of Japan (Adachi, Shinoda, and Izuho 2015), and a much more recent migration of East Asian rice farmers, the Yayoi culture, approximately 2300 cal yr BP (Habu 2004;Hanihara 1991). It is the earlier of the two components that is associated with the production of stemmed points in Japan 16,000-15,000 cal yr BP.
According to the "Out of Japan" model, warming climates following the Last Glacial Maximum (after 18,000-17,000 cal yr BP) invited northward expansion of the people who made the Incipient Jomon points in the Northeast Asian maritime region, who already had developed a successful coastal economy with a marine diet and watercraft. They are assumed to have spread rapidly along the southern coast of Beringia (which has been characterized as a "kelp highway" (Erlandson et al. 2007)) and into the rich marine environments of the deglaciated Pacific Northwest of North America, bringing their distinctive stemmed points with them (Erlandson 2013). 3 In our view, many elements of this model are highly plausible and some of it may be regarded as at least tentatively confirmed by research undertaken during the past two decades. We concur with Erlandson (2013) and others (Davis et al. 2019;Dixon 2001;Erlandson et al. 2007) that Native Americans probably dispersed initially along the Pacific Northwest coast with the aid of watercraft and subsequently spread down the Pacific coast of South America, and that the first occupants of mid-latitude North America probably pursued a coastal economy. Moreover, whatever the biological relationship between the people who made stemmed points in Japan and California 15,000-13,000 cal yr BP, the similarities in their artifacts are striking (e.g., Davis et al. 2019, figure 5;Erlandson and Braje 2011, figure 4).
Given the explosive growth of genetic data in recent years, the time is right to re-evaluate hypotheses regarding the peopling of the Americas with a synthesis of biological and archaeological evidence (e.g., Pinotti and Santos 2020;Raff and Bolnick 2015). In what follows, we address the "Out of Japan" hypothesis with the data and methods of biological anthropologyboth anatomical and geneticand show that the people who made the Incipient Jomon artifacts are an unlikely source for Indigenous Americans. We suggest the source populations for the latter must be sought elsewhere.

Craniometrics
Human cranial variation has a long and, originally, wrought history within the scientific investigation of the peopling of the Americas. In the nineteenth century, anatomists in the United States measured cranial size across Indigenous peoples in the Western Hemisphere with the intent of identifying differences in intelligence compared to Europeans (e.g., Morton and Combe 1839; see early overview by Hrdlička 1919). Craniometrics gradually evolved to focus less on racialized questions and more on understanding the ancestral affinities of human populations. Twenty years ago, human cranial variation joined dental variation and molecular genetics as a key dataset for querying the biological affinities of the Indigenous people living in the Western Hemisphere (e.g., González-José et al. 2005;Herrera et al. 2017;Hubbe, Neves, and Harvati 2010;Hubbe et al. 2020;Kuzminsky, Coonerty, and Fehren-Schmitz 2017;Neves and Pucciarelli 1991).
Initially, craniometric evidence suggested that the earliest people occupying the Americas (< 8000 years ago) differed from later people, with the former having crania that were more anteroposteriorly elongated (e.g., González-José et al. 2005;Neves and Hubbe 2005;Neves and Pucciarelli 1991). In contrast to other lines of evidence, these cranial shape data suggested that the earliest people in the Americas had a stronger biological affinity with people living in the South Pacific rather than East Asia, and perhaps evidenced two waves of migration (e.g., González-José et al. 2005;Hubbe, Neves, and Harvati 2010;Neves and Hubbe 2005;Neves and Pucciarelli 1991).
As larger sample sizes and more nuanced analyses became available, the morphological distinction between early and later Indigenous Americans was increasingly realized to be less obvious or abrupt than originally thought (Kuzminsky, Coonerty, andFehren-Schmitz 2017, 2018). However, the craniofacial shapes of Indigenous populations living in the Americas did change over time, providing a rich dataset for probing the biological affinities and models of dispersion among populations within the Americas (e.g., Herrera et al. 2017;Menéndez et al. 2015).
In terms of understanding the deeper ancestry of Indigenous Americans, four skulls from Quintana Roo, Mexico, add an interesting twist (Hubbe, Neves, and Harvati 2010). These skeletal remains are from people who lived 13,000-12,000 cal yr BP on the Yucatan peninsula (Chatters et al. 2014). Two of the skulls show morphological affinities with North American arctic populations; another has strong similarities to Paleoamericans in South America; and the fourth has similarities to Europeans (Hubbe et al. 2020). As the authors of that study note, "At the very least, it provokes researchers to reevaluate the validity of extrapolations made in the past" (Hubbe et al. 2020, 16).
To infer ancestry from anatomical variation, that variation would ideally be influenced by many different genomic loci and result from only neutral evolution. This is a high hurdle for human cranial variation, given the complex interplay of genetic, non-genetic, and developmental forces that influence its shape, and how intertwined cranial variation is with mastication, thermoregulation, mate choice, etc. (Lieberman 2011). Quantitative genetic-driven studies of human cranial variation enable us to unravel some of this intricacy (reviewed in von Cramon-Taubadel 2014).
Two recent studies are of particular relevance to the use of Indigenous American craniofacial variation to infer ancestry. De Azevedo et al.'s (2017) study of Indigenous Americans found that the observed craniofacial variation accords with a single ancestral source, supporting the "Beringian standstill" model (Tamm et al. 2007). They also found that the variation accrued since that diverse ancestral population deviates from a neutral evolutionary model, indicating that microevolutionary forces have been at play (de Azevedo et al. 2017). Katz, Grote, and Weaver (2017) explored how the shift from foraging to farming may have influenced human craniofacial form globally. Results showed a modest but significant directional effect between populations that rely on the much softer versus tougher diets (Katz, Grote, and Weaver 2017). Given the Katz, Grote, and Weaver (2017) results, it is quite likely that the diversity of subsistence strategies that Indigenous Americans adopted as they migrated across the Western Hemisphere contributed to the microevolutionary forces inferred by de Azevedo et al. (2017). Considering the vast latitudinal range of the Americas, climate likely played a role, too (Menendez 2018;von Cramon-Taubadel 2014), further obscuring evidence of ancestry.
In our view, craniometrics does not provide an adequate basis for identifying the source population(s) of Native Americans.

Dental anthropology
One hundred years ago, Aleš Hrdlička (1920) wrote a seminal paper on shovel-shaped incisors and demonstrated how a morphological crown trait could provide clues on long-term human population history. Specifically, he noted the widespread occurrence of the trait in Native Americans, which differs significantly from the lower frequencies and trait expressions typical of Euro-and African-Americans. 4 Half a century later, Christy G. Turner II (1971) used another trait, a three-rooted lower first molar (3RM1), to develop a model for the peopling of the Americas, 5 which subsequently became a component of the "three-wave model," based on linguistic, dental, and genetic data (Greenberg, Turner, and Zegura 1986).
Aware of the limitations of single traits (or genes) for reconstructing population history, Turner and his students later identified and classified over three dozen crown and root traits that culminated in the Arizona State University Dental Anthropology System (ASU-DAS) (Edgar 2017;Turner, Nichol, and Scott 1991;Scott and Irish 2017). This standardization ushered in a new era of global dental morphology research. Detailed trait descriptions and the availability of plaques for many traits has significantly reduced observer error, a problem in pre-1970s research (Scott, Maier, and Heim 2016;Scott and Pilloud 2019).
Most dental morphological traits are either absent or exhibit a range of expression from slight to pronounced when present. Although twin and family studies show dental traits are under strong genetic control (Hughes and Townsend 2013;Paul et al. 2020;Townsend et al. 2009), they do not have simple Mendelian modes of inheritance like blood group antigens, serum proteins, and other simple genetic markers (cf. Roychoudhury and Nei 1988). Therefore, they have been assumed to be quasicontinuous or threshold traits with polygenic modes of inheritance (Harris 1977;Scott 1973). For that reason, nonmetric dental characters are characterized by "trait frequencies" rather than gene frequencies. Trait frequencies represent the total frequency of presence, but in some instances, trait frequency is defined by "frequency above a breakpoint," and consequently are studied as though they are a presence/absence phenomenon.
Although the identification of major gene effects on crown trait expression met with limited success in early studies (e.g., Kolakowski, Harris, and Bailit 1980;Nichol 1989Nichol , 1990, researchers have demonstrated that shovel-shaped incisors are partly controlled (ca. 19%) by the EDAR V370A allele (Kimura et al. 2009) that has effects on hair, skin, and teeth. Other dental traits, including lower molar cusp number, are also linked to this allele (Park et al. 2012). Increasingly, dental trait evolution is being investigated as part of a pleiotropic milieu (e.g., Hlusko et al. 2018). Although population variation in crown and root traits has been dictated primarily by random factors (genetic drift, founder effect) associated with time since divergence and geographic isolation (Monson, Fecker, and Scherrer 2020;Rathmann and Reyes-Centeno 2020), some traits may be indirectly affected by natural selection acting on variables critical to survival and reproduction (i.e., they can act like "genetic hitchhikers" (Scott et al. 2018a)).
Researchers have recently used genomic data in concert with dental morphological data to reconstruct population history (Hubbard 2012;Hubbard, Guatelli-Steinberg, and Irish 2015;Irish et al. 2020;Rathmann and Reyes-Centeno 2020;Rathmann et al. 2017;Reyes-Centeno et al. 2017). Although there is no one-to-one correspondence in results, these analyses show dental morphological (nonmetric) traits produce high and significant correlations with neutral genomic data. Rathmann and Reyes-Centeno (2020) note some traits are closely in line with neutral genomic markers while other traits show patterns of variation consistent with the action of selection. Irish et al. (2020, 25) conclude: The bottom line then, in conjunction with the recent heritability studies and population analyses, and despite potential concerns (sex dimorphism, selection, etc.), is that dental nonmetric traits actually can and do work well as proxies for neutral genomic data.
Additionally, crown and root morphologic traits are highly resistant to change when populations migrate from one area to another, no matter what new climatic, dietary, and disease factors are encountered. When Europeans colonized the Americas, Africa, and Australia, their dental traits continued to exhibit a Eurodont pattern (Scott et al. 2013), one that provides a stark contrast to the teeth of the indigenous populations of the other three continents.

Samples
Over a span of three decades, C. G. Turner II amassed the world's largest database on human tooth crown and root morphology. His initial interest was the use of teeth to address the peopling of the Americas (cf. Turner 1971), but after collecting data on almost 10,000 Native Americans from the Arctic, North America, and South America, he shifted his focus to Asia and the Pacific. He also collected data on many European populations, while Africa and India received limited attention. His tabulated data and individual data sheets are at the core of this analysis. 6 The age of the samples in the database is highly variable and in large part dependent on what skeletal material was recovered and curated in the various museums on Turner's itinerary. In his provenance forms, materials recovered prior to the advent of radiocarbon dating are noted as recent, historic, or prehistoric; archaeological periods are noted when possible. Where radiocarbon dates were available, the temporal ranges for a sample were noted with more precision ( Table 1).
As the focus on this paper is the biological relationship between the Jomon and Native American populations, only those groups are noted for each time period. First, in the Americas, Turner examined Paleoindians when possible, but they are limited in number and often provide little data due to pronounced tooth wear. The Native American samples used in this analysis range in age from historic to early Holocene. It was not feasible to break North and South American samples into finer age categories. While this would be a serious problem with archaeological materials, it is less significant for teeth. The dentition is under strong genetic control and is highly conserved from a temporal standpoint.
Turner made observations on seven Jomon samples, mostly from Honshu but also from Kyushu and Hokkaido. The Hokkaido sample (n = 136) is noted as Early to Late Jomon, with other samples (n = 587) noted as Middle to Late Jomon or only as Late Jomon. On the computer printouts, he used a combined Jomon sample because they all showed similar frequency profiles (a finding verified in this analysis). The five Ainu samples from Hokkaido and Sakhalin (n = 330) were historic in age. Like the Jomon, the Ainu samples were uniform, so his frequency table combined the five samples into one large Ainu sample.

Measuring biodistance
In biological anthropology, it is commonplace to use distance statistics to evaluate similarities and differences between groups and to make inferences about population origins and relationships (cf. Pilloud and Hefner 2016). The fundamental assumptions underlying biological distance statistics are listed by Scott (1992, 152), beginning with groups with numerous genetically mediated biological differences are distantly related, while groups showing a preponderance of similarities have a recent common ancestor. Traits should be neutral for the most part with founder effect and genetic drift as the primary driving forces in differentiation. Traits under strong selective pressure (e.g., sickle-cell anemia) should not be used to assess affinity. Gene flow between two or more groups results in convergence and can have an adverse impact on distance statistics.
To avoid sampling error, as many traits as possible should be used to calculate pairwise distances. Dissimilarity statistics are based on the premise that a group's distance from itself is "zero" and that small pairwise values indicate closer similarity than large pairwise values. Euclidean and Bray-Curtis distances were used here, but the results from the two methods were highly correlated (r = 0.94), so we present only the results of the Euclidean distances computed using Numerical Taxonomy System software (pc-NTSYS; Rohlf 1993). Distance matrixes are the basis for generating dendrograms or "trees" to provide a visual representation of relative distance. Of the many options available through pc-NTSYS, we use the unweighted pair groups method (UPGMA).

A new method: rASUDAS
Although population differences in dental morphology among world populations are well established (Scott et al. 2018a), only recently has a web-based application been developed to evaluate the ancestry of single individuals within a sample (rASUDAS or R programming in conjunction with the Arizona State University Dental Anthropology System). 7 Using a naïve Bayes algorithm, the application calculates the probability that an individual can be assigned to one of seven major geno-geographic groups: Western Eurasia, Sub-Saharan Africa, East Asia, Southeast Asia & Polynesia, Austral-Melanesia & Micronesia, American Arctic & Northeast Asia, and American Indian. (Given their indeterminate status (see below), data on Jomon and Ainu were not used to derive the East Asian frequencies used in the Bayes algorithm.)

Jomon and Ainu
Our analyses are designed to test the hypothesis that the Native American founder population is derived from the people who made stemmed points in Japan and adjoining areas of Northeast Asia roughly 16,000 cal yr BP. These people are known as the Jomon and they represent an earlier population in Japan that was partially replaced through admixture by the Yayoi ancestors of the modern Japanese people ∼2300 cal yr BP. The northern Japanese island of Hokkaido is inhabited by a population (Ainu) that is more closely related to the Jomon than to contemporary Japanese on the main island (Turner 1976). Our analyses are therefore focused on the relationship between Native Americans and the Jomon/Ainu.
As the relationships of the Jomon and Ainu are the primary focus, we excluded Western Eurasia (Europe, India, North Africa, Middle East) and Sub-Saharan Africa samples from our analysis. In the biodistance analysis and rASUDAS application, we included populations from East and Southeast Asia, the New World, and Pacific. All samples in the biodistance analysis are composites of many groups and most sample sizes are very large (thousands of individuals). Table 2 lists the crown and root traits and their breakpoints for the biodistance and rASUDAS analyses. Table 3 shows the dental trait frequencies for the Jomon and Ainu, along with 14 additional regional groups.
An examination of the 21 crown and root trait frequencies (Table 3) shows Jomon and Ainu are like one another for some characteristics but exhibit distinct differences for others. For many traits, they differ from all other Asian, Pacific, and New World populations. Compared to East Asians, the Jomon and Ainu have: (1) significantly less shoveling and double-shoveling of UI1; (2) more 3-cusped UM2; (3) lower frequencies of UM1 Carabelli's trait; (4) more root reduction of UM2; (5) more four-cusped LM2; and (6) a lower frequency of three-rooted LM1. In some instances, Jomon frequencies are closer to East Asia than they are to the Ainu (e.g., UI1 winging, cusp 6 LM1), but conversely, the Ainu are closer to East Asians for even more traits (e.g., UI1 interruption grooves, UM1 enamel extensions, pegged-reduced-missing UM2, and LM2 root number).
Although a trait-by-trait comparison provides some insights into the affinities of the Jomon and Ainu, in particular the significant reduction of shoveled incisors, pairwise distance values provide an overall summary of trait frequency differences among the 16 groups in the sample matrix.

Biodistance with a focus on Jomon/Ainu
In a distance matrix with 16 groups, there are 110 pairwise distances (Table 4). For mean Euclidean distances, the closest values for both Jomon and Ainu are with Southeast Asia & Polynesia (0.690/0.741). Mean distances from Austral-Melanesia are almost identical for the two groups (0.930/0.986). The average distances to the Americas are substantially larger for the Jomon (1.122) but are like the distance value from Austral-Melanesia for the Ainu (0.973). Relative to East Asia, the Jomon are more distant (0.959) than the Ainu (0.741), likely a reflection of gene flow between mainland Japanese and Ainu populations in recent times.
Euclidean distance values were used to derive the UPGMA tree (unweighted pair groups) shown as Figure 2. The dendrogram exhibits a deep primary bifurcation between East Asians and Native Americans in the lower branches, and between Southeast Asians and Pacific populations in the upper branches. Jomon and Ainu cluster with one another and, excluding Micronesians, are the first branch to separate from Southeast Asian and Pacific populations. While Jomon and Ainu cluster together, the split between the two is deeper than any of the East Asian and Native American branches.
Some general observations are: (1) although Jomon and Ainu are far from being identical to one another dentally, they are more similar to one another than they are to any other Asian, Pacific, or New World grouping; (2) the five Native American groups always cluster together and are aligned with East Asians; and (3) Jomon and Ainu cluster with Southeast Asian and Pacific populations. For Jomon, the second most similar group was Austral-Melanesian, but for the Ainu, the second group was East Asian. For both groups and both distances, they were most divergent from Native Americans.

Individual ancestry (rASUDAS)
The web-based application rASUDAS2 has scroll-down categories where individuals are scored for a series of crown and root traits (minimum of 12, maximum of 26). Once data are entered, the program produces a table with the posterior probabilities of that individual being assigned to one of the five geno-geographic groups noted previously (Western Eurasians and Sub-Saharan Africans were excluded from our analysis).
In Table 5, the frequencies of individuals with a primary assignment to each of the five major groups are listed in columns under each group. For example, 12.5% of North Alaska Iñupiat had the highest probability of assignment with East Asia; 58.9% were classified as American Arctic, 20.3% as non-Arctic American, 5.7% as Southeast Asian, and 2.6% as Austral-Melanesian. Row totals equal 100% (values noted as frequencies in the table, not percentages). For 10 of 12 regions, multiple samples were evaluated. Chi-square values were computed to determine if there were intraregional differences in ancestry assignment. In no case was there a significant difference between regional samples, including comparisons of the three Ainu and four Jomon samples. As this is the first major effort using rASUDAS to evaluate individual ancestry, the consistency of assignments within regions is a significant finding.
In addition to ancestral assignment percentages, there is an additional row that shows mean posterior   probabilities. Every individual has a probability of assignment to a specific group. These probabilities were averaged to provide another measure of group assignment. The two values (per cent assignments and mean probabilities) show the same pattern of ranking but differ in specific values. Figure 3 provides a visual representation of per cent ancestry assignments for the major population groups of Asia, the Pacific, and the Americas.
3.6.1. Americas General observations are: (1) the total assignment of individuals to American Arctic and non-Arctic American is between 75 and 85%; (2) from north to south, the assignment of individuals to American Arctic follows a gradient from 59% to 43% to 23% to 16%; (3) from north to south, the assignment of individuals to non-Arctic American increases from 20% to 32% to 59% to 69%; (4) assignments to East Asia are consistent, ranging from 10.6% to 15.2%; and (5) assignments to combined Southeast Asia and Austral-Melanesians are 4-10%.

Pacific
The assignment of Australians and Melanesians to Austral-Melanesia is 80%, higher than that for any other group. There is a hint of Southeast Asia, but East Asia and the Americas barely register. Micronesia, represented by Guam, shows an almost identical contribution from Southeast Asia (39%) and Austral-Melanesia (39%), with a small number assigned to East Asia (9%) and the Americas (13%). Polynesia, represented by a Hawaiian sample, shows a stronger Southeast Asian component (46%) than Micronesia but still retains a substantial Austral-Melanesian aspect (29%). Percentages are lower for East Asia (13%) and the Americas (13%).

Asia
Southeast Asia shows a relatively low percentage of individuals that are assigned to Southeast Asia (32%), which is not surprising (in most analyses of crown and root morphology, Southeast Asia is intermediate among the dental patterns of Afridonty, Eurodonty, and Sinodonty). Slightly more Southeast Asians are classified as Austral-Melanesian (34%). On some occasions, individuals from this area are classified as East Asian (19%) or Native American (16%). For East Asia, Japan has a plurality of individuals assigned to East Asia (38%), but the two combined American groups reach the same percentage (38%). Some mainland Japanese classify as Southeast Asian (17%) but rarely as Austral-Melanesian (8%). China and Mongolia show a stronger East Asian signal than Japan, where half of the individuals (51%) classify as East Asian. The Native American signal remains strong for China and Mongolia (33%, combined for two New World groups), but fewer individuals are assigned to Southeast Asia (12%) and Austral-Melanesia (4%).

Jomon and Ainu
Although they differ in some respects from the biodistance measures (see above), the results from the rASU-DAS analysis are similar for Jomon and Ainu. For both  . Surprisingly, only 9-14% were assigned to Southeast Asia, the region with the lowest average distance values (Figure 3).

Out of Jomon/Ainu?
The UPGMA tree based on Euclidean biodistance values yields the following clusters: (1) the five Native American groups; (2) the three East Asian groups; (3) the four Austral-Melanesian groups; and (4) the Jomon and Ainu. The tree based on Euclidean distances shows Jomon and Ainu cluster with Southeast Asia and the Pacific. The analysis of individual ancestry through rASUDAS2 provides a different perspective on Jomon/ Ainu links. Over 40% of Jomon individuals were assigned to Austral-Melanesian, with the second most common assignment being American Arctic (27%). More individuals are assigned to East Asia (15%) than to Southeast Asia (9%), a finding at odds with their smallest genetic distance to Southeast Asia. The mean posterior probabilities show a slightly different picture. While Austral-Melanesian still has the highest average posterior probability (0.312), the values for American Arctic (0.231), East Asia (0.187), and Southeast Asia (0.190) are almost identical. The only group that does not match the Jomon dental pattern is non-Arctic American (5.5%; 0.080 average). And compared to other groups, the Jomon are dentally heterogeneous. For individual ancestry assignment, the Ainu parallel the Jomon to some extent. Fewer are classified as Austral-Melanesian (28%), while more are classified as American Arctic (31%). Again, more are assigned to East Asia (20%) than to Southeast Asia (14%), with very few classified as non-Arctic American (7%). The average posterior probabilities are even more uniform for the Ainu than for the Jomon, with (3) the non-Arctic American component for both the Jomon (5.5%) and Ainu (6.8%) is the smallest for either group; (4) in contrast to the Jomon and Ainu, Native American samples show a small (< 10%) component of Southeast Asian/Austral-Melanesian combined, and a consistent East Asian component (10-15%); and (5) consistent with a long period of isolation and differentiation from a late Pleistocene common ancestor, 75-85% of all Native Americans are classified as either Arctic American or non-Arctic American. Given the significant contrast between the Jomon/Ainu and all Native American lineages, dental morphology does not support any ancestral relationship between the earlier peoples of the Japanese archipelago and the populations of the New World from its northernmost to southernmost extreme.
Anthropologists have long been challenged to determine the place of the Ainu in the broad web of human history. Their teeth defy simple or obvious classification. The Jomon present the same conundrum: they have elements of mainland Asia, Southeast Asia, Austral-Melanesia, and American Arctic. The only element they both lack is the hyper-Sinodont dental pattern displayed by the Native Americans of mid-latitude North America and South America (i.e., outside the Arctic) (Scott et al. 2018b). The dental morphological data suggest that the biological contribution of the Jomon to the settlement of the Americas is negligible or nonexistent.

Human genetics
Human genetics provides an even more direct measure of biological relationships among present and past populations than teeth, but until recently it was possible only to examine the genetics of the Ainu and other living groups, including living Native Americans. The development of techniques for recovering and analyzing ancient DNA (aDNA) in human skeletal remains dating to tens of thousands of years ago transformed paleoanthropology and made it possible to test the hypothesis that Native Americans are derived from the people who made stemmed points in Japan 16,000-15,000 cal yr BP with genetic data. We are limited only by the recovery of pertinent skeletal remains with adequate preservation of aDNA.

Native American paleogenomics
The ancestry of the First Peoples of the Americas lies in two major sources. The first source derives from a population that separated from the ancestors of East Asians (represented by present-day Han) approximately 30,000 cal yr BP (95% CI 26.8-36.4 ka) Sikora et al. 2019). This population subsequently diverged into an "Ancient Palaeo-Siberian" lineage and an "Ancient Beringian" lineage around 24,000 cal yr BP (95% CI 20.9-27.9 ka), with the latter represented by Upward Sun River 1 (USR1) (Moreno-Mayar et al. 2018). 8 About 20,000 cal yr BP, gene flow from a second population with west Eurasian roots that currently is best represented by the 24,000-year-old Mal'ta child contributed ancestry representing 18.3% (95% CI 9.8-20.3%) of the USR1 genome (Raghavan et al. 2014;Sikora et al. 2019). It is this combination of ancestral Mal'ta and ancestral East Asian that comprises the majority of the Native American gene pool, although where and how these demographic scenarios played out during and after the last glacial maximum (LGM) is a matter of ongoing debate Ning et al. 2020;Sikora et al. 2019;Yu et al. 2020).
To test the hypothesis that post-LGM Japan is the source population for Native Americans (e.g., Davis et al. 2019), it is essential to focus our analysis on the people who occupied Japan before 2300 cal yr BP, rather than the contemporary mainland Japanese population, which is derived from the admixture of early Jomon peoples and Yayoi migrants from mainland Asia approximately 2300 cal yr BP. While the Jomon made a limited genetic contribution to the modern Japanese population (i.e., < 20%; Kanzawa-Kiriyama et al. 2017), they share more ancestry with living Ainu people of northern Japan and Ryukyuans of the southern Okinawan Islands (e.g., Yuasa et al. 2015).

Maternal lineages (mtDNA)
Published Jomon mitochondrial DNA sequences are predominantly from Hokkaido, the northernmost island and the environment most amenable to ancient DNA preservation. These individuals, spanning a range of nearly 6000 years (Adachi, Shinoda, and Izuho 2015, 413;Adachi et al. 2011), indicate that the Jomon were represented by a set of maternal lineages that are not found in contemporary or ancient Native Americans, nor are direct ancestral or descendant clades to Native American lineages (Adachi et al. 2011(Adachi et al. , 2013Horai et al. 1989;Kanzawa-Kiriyama et al. 2013;Perego et al. 2009Perego et al. , 2010. The most common mitochondrial haplogroup identified from Hokkaido Jomon remains is N9b (at ∼65%), present in contemporary Siberian populations but rare in East or Southeast Asia. Haplogroups M7a, G1b, and D4h2 have also been identified (Adachi, Shinoda, and Izuho 2015, table 28.3;Kanzawa-Kiriyama et al. 2013, appendix 1). Limited sampling of a small number of human remains from Honshu further expands this list to D4b2, as well as additional haplotypes of N9b and M7a (Adachi et al. 2013;Kanzawa-Kiriyama et al. 2013;Mizuno et al. 2020;Takahashi et al. 2019). Of particular interest for this discussion is macro-haplogroup D4, which is widely distributed and diversified across Asia.
Daughter clades of haplogroup D4 are represented in Native American populations. These are Holocene-aged haplogroups D2a and D4b1a2a1a, limited to Arctic populations of North America and Chukotka (Tackney et al. 2019), and the rare Native American founding haplogroup D4h3a (Perego et al. 2009). While the first two haplogroups are only distantly related to the Jomon types, haplogroup D4h3a is more intriguing. This haplogroup has been typed at low frequency along the Pacific coast of North and South America in ancient and contemporary individuals, though occasionally is found farther inland, most notably in the Clovis-associated Anzick-1 child (Perego et al. 2009;Posth et al. 2018;Rasmussen et al. 2014). The most closely related haplotype to D4h3a is a single D4h3b lineage from the Shandong province of China (Perego et al. 2009), followed by D4h2, which is present not just in the Jomon, but the Ulchi of Khabarovsk Krai (Kanzawa-Kiriyama et al. 2019). Both, however, are multiple mutational steps away from D4h3a and therefore temporally quite divergent. Sato et al. (2010) characterized the molecular diversity of the ABO blood group system in Jomon/Epi-Jomon and Okhotsk people from Hokkaido. These molecular markers of the ABO system showed slight differences between the Jomon/Epi-Jomon and either the Okhotsk or modern Japanese. In addition, the Jomon/Epi-Jomon showed some similarity to populations in eastern Siberia, but whether this was the result of ancestral-descendant relationships or gene flow could not be determined in these samples.
In a larger study of Japanese Y-chromosome diversity in contemporary Japan, Watanabe et al. (2019) identify the D1b Y-lineage as deriving from early Jomon because it is at high frequency in Japan, but effectively absent from the rest of East Asian populations, except Tibetans. This paternal lineage represents ∼35% of Y-lineages in Japan (Jinam, Kanzawa-Kiriyama, and Saitou 2015;Kanzawa-Kiriyama et al. 2017). Watanabe et al. (2019) also identify a significant Jomon population decrease near the end of the Jomon period coincident with a climatic change, as well as the introduction of Yayoi migrants. This population bottleneck, inferred from Y-chromosome variation, was also detected by Karmin et al. (2015) in a global analysis of Y-lineage diversity.

Nuclear genomes
Complete nuclear genomes have been reported from several Jomon individuals. Kanzawa-Kiriyama et al. (2017) published low depth of coverage (∼0.03×, > 115 million base pairs) genome sequences from two 3000-cal-yr-BP Jomon individuals from the Sanganji Shell Mound in the Tohoku region of Honshu. Despite the low depth, these partial genomes confirmed that these Sanganji Jomon individuals carried mitochondrial haplogroup N9b, that the Jomon diverged prior to an East Eurasian-Native American split, and that the closest living populations to the Jomon are the Ainu, then the southern Ryukyuan, and then the Honshu Japanese. Kanzawa-Kiriyama and colleagues estimated that the genetic contribution of the Jomon to contemporary mainland Japanese peoples was low, with contemporary Japanese genomes derived primarily from multiple later groups admixing with indigenous Jomon.
Kanzawa-Kiriyama et al. (2019) obtained nuclear genomes from two late Jomon individuals dating to between 3500 and 3800 cal yr BP from the Funadomari site on Hokkaido. These were sequenced at a much greater depth of coverage than the Sanganji Jomon individuals (preservation, as expected, was better for the Hokkaido remains). Interestingly, both individuals, a male (F5) and a female (F23), share the CPT1A mutation with North American arctic populations that has been attributed to adaptation to a high-fat diet.
Whether this is a homologous trait or an example of convergence cannot be determined with these data. In a phylogenetic analysis of the F23 individual, Eurasians, and Native American groups, a deep split of F23 from ancient East Eurasians is again suggested, prior to the divergence of Native Americans from an East Asian background. The authors offer a wide confidence interval of 38,000-18,000 cal yr BP that partially overlaps the 95% confidence interval of 36,400-26,800 cal yr BP for the population split time calculated in Sikora et al. (2019).
In addition to precluding Jomon from proximate Native American ancestry, these results reconfirm the view that the Jomon were isolated from continental East Eurasian populations for a considerable period (Yang et al. 2020). The Funadomari Jomon had a shared genetic affinity with contemporary Ulchi, but not the 7700-cal-yr-BP Neolithic East Asian Devil's Gate individuals (Siska et al. 2017), the supposed primary ancestors of the Ulchi. This implies that Jomon gene flow with coastal East Asians was a much more recent development. Kanzawa-Kiriyama et al. (2019) calculate a small effective population size for the Hokkaido Jomon (though precluding consanguinity), and estimate Jomon ancestry in Honshu Japanese, Ryukyuan, and Ainu at 9-16%, 27%, and 66%, respectively. The next closest contemporary populations were Korean, aboriginal Taiwanese (Fu et al. 2013), as well as earlier than the divergence between contemporary and ancient Tibetans, and earlier than the divergence between East Asians and the ancestors of Native Americans (see also Yang et al. 2020; Figure 5).
Gakuhari and colleagues argue that the genome of IK002 represents an individual from the basal lineage for all East Eurasians, a population of the earliest migrants from Southeast Asia toward East Asia (but see Zhang and Fu (2020) for an alternative view of the directionality of these migrations). The observed genetic affinity between IK002 and an 8000-cal-yr-BP Hòabìnhian hunter-gatherer (Laos) indicates a direct link between Jomon and Southeast Asians (McColl et al. Kanzawa-Kiriyama et al. (2019), the contemporary Ainu are again deemed the closest living population to the Jomon, with approximately 79% ancestry, followed by main island Japanese, Taiwanese Ami and Tayal, Ulchi, and Nivhk (all coastal East Asian populations).

Discussion
Despite limited genomic information on early Jomon individuals, several inferences are strongly supported by current analyses. First, neither the mitochondrial, Y-chromosome, or nuclear genomic sequence data suggest a direct relationship between Jomon populations and early Native Americans. This is particularly clear in the distribution of maternal and paternal lineages, which do not overlap between the early Jomon and American populations.
Second, prehistoric Jomon populations were genetically heterogeneous. Subdivision and differentiation among these early groups appears to have been the norm. For example, Kanzawa-Kiriyama et al. (2013) summarized mtDNA lineage diversity among several regional Jomon sites. Kanto Jomon on "mainland Japan" were by far the most diverse, exhibiting at least 11 distinct mitochondrial lineages, while the Hokkaido Jomon possess only four mtDNA lineages, and the geographically intermediate Tohoku Jomon only three. Only two mtDNA lineages are common to all the Jomon samples, M7a and N9b. Even within lineages at a single site sequence, diversity is observed.
This level of differentiation and diversity is entirely consistent with the dual structure model of Japanese origins (Hanihara 1991;Jinam, Kanzawa-Kiriyama, and Saitou 2015), with contemporary Ainu and Ryukyuan populations occupying the northern and southern islands, respectively, of the Japanese archipelago most closely resembling Jomon individuals (Jeong, Nakagome, and DiRienzo 2016;Kanzawa-Kiriyama et al. 2019). The Ainu subsequently experienced gene flow with the neighboring coastal Okhotsk culture (Adachi et al. 2018). Populations on the main island of Japan experienced multiple incursions of migrants and admixture events, such that the Jomon contribution to contemporary main island Japanese genomes reflects less Jomon ancestry (∼10-20%) and more the influx of Yayoi rice farmers after 3000 cal yr BP (Adachi et al. 2013Gakuhari et al. 2020;Kanzawa-Kiriyama et al. 2013Schmidt and Seguchi 2014).
It should come as no surprise that ancient Jomon people exhibited genetic variation different than that seen in contemporary populations of the region. This has been observed not only throughout South and East Asia (Yang et al. 2020), but in South America (Llamas, Harkins, and Fehren-Schmitz 2017) and Europe (Fu et al. 2016). Contemporary, and even archaeologically recent, populations in any region may be poor indicators of patterns of genetic variation in past populations of the deeper past of the same regions.
Estimates of an early differentiation of the Jomon population may also signal isolation from other continental (East Asian) populations until the later Yayoi migrations. The observed diversity and population structure in the Jomon individuals studied to date suggest that additional genomic analyses will be required to fully document Jomon origins and interactions with other regional populations in the late Pleistocene. Additionally, the small population size of hunter-gatherer populations like Jomon, coupled with isolation and population differentiation across the archipelago, is consistent with stochastic effects on the genome. Such stochastic effects (genetic drift) are equally powerful modifiers of morphometric phenotypes (Fukase et al. 2012;Morita et al. 2012;Schroeder and von Cramon-Taubadel 2020;von Cramon-Taubadel 2019). Such patterns of diversity and differentiation among geographically dispersed Jomon might well result in the expanded variation observed in Jomon dental morphology and the apparent affinity of these morphological phenotypes with a variety of early ancestral and subsequent descendant groups.

Discussion and conclusions
We have addressed the hypothesis that the First Peoples of the Western Hemisphere are derived from a population living in the Northeast Asian maritime region, including northern Japan, at the end of the LGM or about 16,000 cal yr BP. The hypothesis is based on striking similarities in a specific category of stone artifacts (stemmed points) made by (1) the people who occupied Japan and adjoining areas of mainland Asia 16,000-15,000 cal yr BP (known archaeologically as "Incipient Jomon") and (2) early Native American people living in the western regions of mid-latitude North America ("Western Stemmed Point Tradition").
We tested the hypothesis with the data and methods of biological anthropology, using a large body of data on teeth, which are proxies for neutral genomic data, and the growing body of pertinent genetic data itself (paleogenomics). We found that (1) quantitative biodistance measures using large samples of teeth and based on 21 crown and root traits show that the Jomon (and Ainu) cluster with Southeast Asian and Pacific groups; they are most divergent from Native Americans; (2) an analysis of individual ancestry through rASUDAS showed that roughly 40% of the Jomon sample were assigned to Austral-Melanesians and a significant percentage (24%) were assigned to American Arctic, but that only 7% were assigned to non-Arctic Native American (the Ainu sample exhibited a similar pattern); and (3) none of the genetic data (mtDNA, Y-DNA, or nuclear genome) suggests a direct relationship between the Jomon and Native Americans.
Our analyses are limited by the fact that available samples of both teeth and ancient DNA for the Jomon population are less than 10,000 years old, i.e., do not antedate the early Holocene. We assume that they are valid proxies for the Incipient Jomon population or the people who made stemmed points in Japan 16,000-15,000 cal yr BP. At the same time, we recognize that changes in the genetics of the Jomon population must have occurred during the 6000-8000 years that followed. In addition to genetic drift over the course of several hundred generations, the Jomon population could have been subject to significant gene flow from one or more areas of mainland Asia during the millennia between the end of the LGM and the early Holocene.
On the other hand, as described above, we found no evidence of links between the Holocene Jomon population and Native Americans in the dental and paleogenomic data (with the exception of the connection between the Ainu and Arctic Americans, discussed further below). If Native Americans were derived from the people who made the Incipient Jomon artifacts, only more-or-less complete replacement of the population of Japan between 16,000 and 9000 cal yr BP could account for this pattern. There is no indication of a wholesale replacement of the population of Japan during 16,000-9000 cal yr BP. Instead, multiple lines of evidence suggest both biological and cultural continuity in the Jomon population (e.g., Adachi, Shinoda, and Izuho 2015;Nakazawa 2017), which, as discussed earlier, appears to have diverged from its own Asian source population before the LGM as a lineage separate from that of the East Asian lineage of the Native American founder group ( Figure 5; Gakuhari et al. 2020, figure 4;Ning et al. 2020;Yang et al. 2020, figure 2A). 10 We conclude that the people who made stemmed points in the Northeast Asian maritime area during the millennia following the LGM are an unlikely source population for the peoples who spread through mid-latitude North America and South America after 16,000-15,000 cal yr BP. In our view, the Incipient Jomon population represents one of the least likely sources for Native American peoples of any of the non-African populations considered in our analyses.
The apparent link between the Jomon and at least some portion of the American Arctic population, warrants a comment. The pattern is apparent in the dental anthropological but equivocal in the human paleogenomics data and apparently indicates some gene flow between the two populations after the settlement of mid-latitude North America and South America had taken place (15,000-13,000 cal yr BP). We suggest that it probably represents one of several population movements/contacts between Beringia (or the Bering Strait region) and Northeast Asia that took place after the divergence of Native American Arctic and non-Arctic populations (e.g., Adachi et al. 2009;Sicoli and Holton 2014;Sikora et al. 2019;Tamm et al. 2007).
Finally, we note that there are archaeological remains in Hokkaido that may be related to the peopling of the Americas, but do not represent the people who made the artifacts assigned to Incipient Jomon. There are artifact assemblages in Hokkaido dating as early as ∼25,700-22,500 cal yr BP containing microblade cores (including Yubetsu type) that are similar to those found in the interior of Siberia and later in Beringia (Buvit et al. 2016;Gomez Coutouly 2018). Although the lineage associated with these assemblages and its relationship to Indigenous Americans is unknown, the presence of similar artifacts in eastern Beringia suggests a possible, if not likely, role in the origin of the latter (e.g., Graf and Buvit 2017;Potter et al. 2017). The early microblade assemblages in Hokkaido appear unrelated to Incipient Jomon, but rather reflect an LGM interior adaptation at a time when Hokkaido was connected to Sakhalin and the Asian mainland (but not Honshu) due to lower sea level (Buvit et al. 2016 Turner (1967) knew 3RM1 was common in Eskimo and Aleut populations (e.g., 25-45%). When he shifted his attention to western U.S. Indian samples, he found the frequency of 3RM1 to be much lower (ca. 6%). At this point in time, he had not studied samples of Athapaskans and Northwest Coast Indians, but he did have access to radiographs from a small Navajo sample; their frequency of 3RM1 fell between those of Eskimo-Aleuts and American Indians (9/33 = 27.3%). This finding, limited to a single trait, was congruent with the model Joseph Greenberg was developing for Native American languages. That is, Greenberg proposed three major Native American language groups -Amerind, Na-Dene, and Eskimo-Aleutserved as the foundation for linguistic variation in the New World. These two converging lines of evidence led Turner to propose a tentative three-wave model for the peopling of the Americas. 6. Following Turner's death in 2013, the senior author visited his home in Tempe; with the permission of his daughter Korri Turner and wife Olga Pavlova, he initiated the "C.G. Turner II Legacy Project." This involved several visits to Tempe and many hours of scanning computer printouts and data sheets. Although ASUDAS included 38 dental traits, Turner focused on '29 key traits.' Hundreds of printouts had the class frequency distributions for these 29 traits in both individual samples and larger samples that combined two or more smaller samples. For example, he made observations on five Japanese samples, but these were combined to form one large sample given the internal consistency of trait frequencies. In his paper files, Turner had amassed observations on over 23,000 individuals from all over the world with an emphasis on the Americas, Asia, and the Pacific. During the 1980s and 1990s, he employed a technical assistant who entered all these data, initially on large tape reels and then on 5 ¼ inch floppy disks. Unfortunately, these electronic files were not kept up to date and when GRS discovered the floppy disks, they had suffered 'disk rot' and were no longer usable. Given that, all the individual data sheets were scanned as PDFs to provide a permanent, albeit imperfect, record of Turner's paper files. In this study, printouts of sample frequencies and individual data sheets both come into play in ancestry assessment, with a focus on Jomon and Ainu samples from Japan. 7. This method, referred to as rASUDAS (i.e., r programming in conjunction with the Arizona State University Dental Anthropology System), was developed by David Navega and Joao Coelho at the University of Coimbra, Portugal. The developers wanted the application to be freeware so the link to the beta-version is http:// osteomics.com/rASUDAS/. In this study, we have access to the most recent version developed in July 2020, http://osteomics.com/rASUDAS2/. Details for rASUDAS can be found in Scott et al. (2018c Reich 2018, 176-181;Skoglund et al. 2015). The most parsimonious explanation of this pattern is the East Asian source population, which reflects a significant and early contribution from Southeast Asia (e.g., Yang et al. 2020). 9. Gakuhari et al. (2020) highlight an apparent discrepancy between the genomic results reported thus far and the archaeological model(s) for the spread of microblades during the Upper Paleolithic in Siberia. The lack of genetic affinity between IK002 and the Mal'ta child, who is hypothesized to represent the Northern migration route of Upper-Paleolithic populations into East Asia, needs to be rectified with the arrival of microblade technology in Hokkaido and Honshu during the LGM. 10. While this paper was "in press," Cooke et al. (2021) published a paper in Scientific Advances (see References) that challenged the "dual structure" model for the peopling of Japan, concluding that the modern Japanese population reflects a "tripartite origin." Their conclusion was based on analyses of 12 ancient genomes, including Jomon samples dating as early as ∼8800 cal yr BP (Cooke et al. 2021, table 1). Their results lend additional support to our conclusions regarding the biological relationship between the ancient Jomon and Indigenous Americans, i.e., that the former are an unlikely source for the latter. Their conclusion regarding a tripartite rather than a dual origin for the modern Japanese population does not pertain to the relationship between the ancient Jomon and Indigenous Americans.