Because all human populations present a high degree of internal morphological variability, previous studies based on early Americans' cranial morphology were unable to completely rule out sampling anomalies or outliers as an explanation for the results they generated. Consequently, several specialists have questioned whether the peculiar cranial pattern described above for the first South Americans actually expressed the central tendency of this population (4, 5). This controversy could be addressed by the use of large samples of early Americans, if available.

Here we use a sample of 81 Late Pleistocene/Early Holocene specimens from the region of Lagoa Santa, central Brazil (described in Table 1) to explore the extracontinental morphological affinities of the first Americans. This unique sample results from a long-term research project led by W.A.N. involving a major program of curating and dating museum material, and fieldwork in Lagoa Santa. The Lagoa Santa karst, located near the center of the State of Minas Gerais (Fig. 1), has been known since the 1840s (6) as a key area for the study of the first Americans.

The first professional archaeological excavations in Lagoa Santa were carried out by Wesley Hurt and Oldemar Blasi in 1956 (9). They were unable to find any association between human and megafauna remains in the seven rock-shelters excavated by them at Cerca Grande, the largest limestone outcrop in the region. However, many years after their fieldwork, two radiocarbon dates were reported by them (10). These two dates (9,720 ± 128 and 9,028 ± 120), obtained in Rock-Shelter 6, were the first direct evidence that a large number of human skeletons found in that area, including those uncovered by Hurt and Blasi in Cerca Grande, could be in fact of final Pleistocene/Early Holocene age.

The next contribution toward an improved chronological framework for the Lagoa Santa human skeletal remains came from the excavations of Lapa Vermelha IV, in the mid-1970s (11). The now famous “Luzia” skeleton was uncovered by the excavators 12 m below the surface of the rock-shelter, in a sedimentary deposit estimated by ¹⁴C dates on charcoal to be of final Pleistocene age (12).

Since the work of Lund in Sumidouro Cave in 1842/1843 until the mid-1970s when Luzia was found, at least 250 ancient human skeletons were recovered in Lagoa Santa by naturalists, amateurs, and professional archaeologists. Because the majority of the excavations were reported in Portuguese in local publications, few North American and European physical anthropologists were aware of this important material. For the same reason, until recently, none of these human skeletons had been directly dated by accelerator mass spectrometry to confirm their antiquity.

Because of the importance of this unique material to the debate on the settlement of the New World, W.A.N. has directed an extensive operation since 1994 to generate a reliable chronological framework for the human skeletons of Lagoa Santa. This operation has involved four complementary lines of action: dating directly as many human skeletons as possible by accelerator mass spectrometry; generating minimum dates for some of the human skeletal collections by dating cristaline calcite layers sealing archaeological and paleontological deposits; excavating new archaeological sites to increase the number of human specimens and to understand the details of the local stratigraphic sequence; and visiting classical sites looking for remnants of the original sediments to reconstruct the original archaeological sequences. Few specimens could in fact be directly dated by accelerator mass spectrometry, because absence of collagen in the bones is the norm in Lagoa Santa.

In conjunction, these four lines of research have allowed for the construction of a very solid chronological framework for the specimens used in this work. Table 2, which is published as supporting information on the PNAS web site, presents the absolute dates obtained for 22 human skeletons uncovered from 11 classical sites in Lagoa Santa. Most of the specimens are dated to Early Holocene, specifically between 8.5 and 7.5 kyr. Because these specimens can be assumed to represent a random sample of the total universe of already institutionalized human skeletal remains from Lagoa Santa, they can be taken to represent the burial distribution along time in the region as a whole.

A few human skeletal remains dated to Late Holocene (last 2.0 kyr) were also found in the rock-shelters of Lagoa Santa (data not shown). This late material, very fragmented, was found, however, on the surface of the shelters (sometimes deposited in urns) and not in stratigraphy, as the early ones, and can be easily sorted out from the Early Holocene specimens by means of taphonomical characteristics.

Because of its homogeneity, the burial pattern of the early settling of Lagoa Santa can also be used as a complementary indicator of the chronology of the human remains uncovered when neither stratigraphic control nor direct methods of dating were available (13). In general, the corpses were deposited in hyperflexed position in very shallow graves topped by small blocks of limestone or quartz. Sometimes, these blocks were also used to cover the lateral walls of the grave pits. A small hearth was always established adjacent to the top of the pit, and burning charcoals were thrown into the pit, before sealing. Secondary burials were also popular in ancient days in the region. Red pigment was vastly used in these cases. Because most naturalists and amateurs described the burial conditions of the human skeletons they recovered in Lagoa Santa, these descriptions can also be used, as a last resource, to help to contextualize the old material in time.

The chronology assumed for the Lagoa Santa skulls used in this work is also supported by what is known about cultural continuity and change in the region. Humans were present in Lagoa Santa earlier than or at least as early as Clovis in North America (14). However, only two skulls in the studied sample date back to the final Pleistocene. Dry climatic conditions apparently did now allow for a real human settlement within the karst of Lagoa Santa during that period. As it is widely known, karsts in general are rich in subterranean streams but poor in surface water. During the last two millennia of the Pleistocene, humans were not formally buried. Apparently, when a group was crossing the area or sporadically exploring it, eventual casualties were expediently treated: the corpse was simply laid down in extended position in a natural protected niche within caves and cavities. This was the case of Luzia (13) and possibly Confins Man (8).

From ≈10.0 to 7.5 kyr, more humid conditions allowed for an intense settlement within the karst itself, with the great majority of the local rock-shelters already occupied by 9.6 kyr ago. For still unknown reasons, formal burying in the rock-shelters started only around 8.5 kyr ago and would be intensively practiced only until around 7.5 kyr ago. Except for this, the other elements of the local lifestyle and material culture did not change significantly. This phase of intensive use of the karst by humans was followed by what is now known elsewhere in South America as the “archaic gap.” Also because of drier conditions, humans almost completely abandoned the region during the Middle Holocene. Again, the rock-shelters were occupied only sporadically, as in the final Pleistocene, and no human burial dated to between 7.0 and 3.0 kyr has ever been found in Lagoa Santa.

With a local increase in humidity during the Late Holocene, the Lagoa Santa Karst was again intensively occupied by human groups, this time with a very different material culture and settlement-subsistence pattern when compared to the Paleoindians. Slash-and-burn agriculture and pottery making were part of the cultural repertory of the newcomers. Their dead were buried mainly in urns, some of them deposited on the surface of the rock-shelters.

Fig. 2 presents the positions of all series included in the analysis when plotted in two-dimensional graphs defined by principal coordinates 1 and 2 (PCo1 and PCo2). The amount of information contained in each axis is also presented in the figure. In Fig. 2 A (males, size and shape), the sample from Lagoa Santa shows its strongest morphological affinity with Tolai (a Melanesian sample from New Britain), followed by the remaining Australian series. Fig. 2B (males, shape alone) shows a similar topology, with Lagoa Santa again close to Tolai. Fig. 2C (females, size and shape) places the early Americans very close to Tolai. Both series occupy the region of the graph dominated by Australo-Melanesians. This pattern repeats itself in Fig. 2D (females, shape alone). In the four graphs the reference samples taken from Howells' database are, in general, grouped in accordance with geography, as found by him in his seminal studies (3, 15). The two Brazilian late archaic samples show a very different cranial morphology when compared with the early South Americans. The samples occupy a position in the multivariate space close to East Asians and/or late Amerindians.

To consider the individual variability of the Lagoa Santa sample, principal components analyses were applied to the same set of 55 best preserved skulls used in the principal coordinate analyses, and to the centroids of Howells and late archaic Brazilian series. By doing this, we were able to assess the morphological dispersion of the individual skulls of Lagoa Santa in relation to the central tendencies of the world comparative populations. The results (Fig. 3) show a tendency for late Americans and Asian centroids to remain outside of the 95% confidence ellipses of the individual dispersion of Lagoa Santa. Fig. 3A (males, size and shape) shows that, whereas most of the world series fall within the dispersion of Lagoa Santa, none of the five late American samples used are part of the Paleoindian dispersion. When size effect is corrected (Fig. 3B), fewer series can be found within the Paleoindian dispersion. These are mainly Australian and African samples, although some Polynesian (Guam and North Maori) and American series (Eskimo and Base) can be found in the extremity of Lagoa Santa's 95% confidence dispersion. Fig. 3C (females, size and shape) shows that the majority of the series within the Paleoindian dispersion are Australo-Melanesians and Africans. With the exception of Eskimo, no other late American series are found within Lagoa Santa's female dispersion. The same can be said when size is corrected (Fig. 3D).

Two different hypotheses can be proposed to explain the morphological differences observed between early and late Native South Americans (27). One is a local microevolutionary process that transformed, in situ, the Paleoamerican morphology into that prevailing today among Native Americans. The other is that the Americas were successively occupied by two morphologically differentiated human stocks, with the Paleoamerican morphology entering first.

We believe the second hypothesis is more plausible for three reasons: first, it would be very unlikely that the same evolutionary event (directional morphological change) happened in the Americas and in East Asia in parallel at approximately the same time (the parsimony principle) (28); second, because in South America, at least, the transition between the two morphological patterns was, as far as we know, abrupt (29); and third, cranial morphology has recently been shown to respond adaptatively only to extreme environmental conditions, being therefore much less plastic than originally thought (30). No transoceanic migration is necessary to explain our findings, because Paleoamerican-like humans were also present in East Asia during the final Pleistocene (31–35) and could perfectly well have entered the New World across the Bering Strait. A final solution to this dilemma will depend of course on a better understanding of what was happening in North America at the same time.

Recent archaeological data can be used to support a dual occupation of the New World, either directly or indirectly. Dixon (36), for example, analyzed the diversity of the projectile points found in the earliest sites of North America and concluded that two different and independent cultural traditions (or cultures) entered the continent in the final Pleistocene. According to Dixon, bow-and-arrow technology was brought to the Americas only by the second tradition, because the atlatl was the primary hunting weapon of the first.

A pre-Clovis occupation of the New World, as strongly suggested by recent findings in South America (37, 38), would not strictly be requisite to accommodate our findings. However, it would make them more plausible. The deeper the chronology of the settlement of the Americas, the more plausible the entrance of humans exhibiting a more generalized cranial morphology in the New World.