This essay considers the first human colonization of South America from an interdisciplinary perspective, including bioanthropological, genetic, archaeological and biogeographical evidence. Our objectives are twofold. First, we briefly review models that have been proposed for the peopling of the New World, with special reference to the southern hemisphere, including the Clovis First/Single Origin (Hrdlička, 1937; Haynes, 2002), Tripartite (Greenberg et al., 1986), and Dual Migration (Neves & Hubbe, 2005) models. Second, we review some key results from bioanthropology, genetics, and archaeology and their implications particularly regarding the peopling of South America. Models based primarily on data from a single discipline and/or methodological approach are perhaps less likely to advance our understanding of initial human migrations. Our intention here is to build on previous integrated analyses (e.g., Neves et al., 2003; Dillehay, 2008; Fagundes et al., 2008; Goebel et al., 2008; González-José et al., 2008) and to provide a broad critique and a simplifying perspective.
The Clovis First/Single Origin Model
For decades the prevailing view among scholars interested in the peopling of the Americas has been that Paleoindians arrived in the New World via Beringia from north-east Asia during Clovis times around 11,500 years ago, and that they were the direct ancestors of present-day Amerinds (Dillehay, 1999; Dixon, 2001; Adovasio & Page, 2002; Haynes, 2002; Meltzer, 2004; Goebel et al., 2008). This perspective of the first peopling of the Americas is termed the Clovis First/Single Origin hypothesis, which was first proposed by Ales Hrdlička (1937), a major participant in the typology and race debates in biological anthropology during the early part of the 20th century (Blakey, 1987).
After the discovery of “Folsom Man” in the 1920s, Hrdlicka vigorously refuted all claims for any human presence more than 4000 years old in the New World. However, the discovery of the Clovis site in 1932 turned him into one of the leading advocates in revising the perception of the first humans entering the New World and he conceded that they may have arrived in late Pleistocene times. When Clovis projectile points later appeared at numerous sites in North America, often associated with the bone remains of extinct mammals such as mammoth (dated by radiocarbon means to ca. 11,000 years ago) the idea of Clovis Paleoindians having been the first Americans was born. The Clovis First/Single Origin model was later modified to fit the newly devised chronological and cultural schemes for the initial peopling of the Americas (e.g., Haynes, 2002). More recently, archaeological evidence from various sites scattered throughout the Americas and dating from before Clovis times (e.g., Dillehay, 1997; McAvoy & McAvoy, 1997; Adovasio et al., 1999; cf. Goebel et al., 2008), has overturned the “Clovis-First” paradigm, establishing that the first humans arrived in North and South America in pre-Clovis times (c.f., Bryan, 1986; Dillehay, 1997, 2000; Adovasio & Page, 2002; Stanford & Bradley, 2002; Meltzer, 2004, 2006; Goebel et al., 2008).
Shortly after the discovery of the Clovis site, the question arose as to how early hunters had reached the mid-latitudes of North America during the Wisconsin Glaciation (between 25,000 and 10,000 years B.P.), at a time when the northern half of the continent was thought to be buried beneath two extensive ice sheets called Laurentide and Cordilleran. In 1933, W.A. Johnston proposed the existence of an ancient passage between the two glaciers, later termed the “ice-free corridor” by Antevs, which stretched from the Yucon to northern Montana (Jackson et al., 1997). New geological and archaeological information has doubted the existence of this “corridor” (Jackson et al., 1997; Mandryk et al., 2001; Clague et al., 2004), however, leading researchers to consider other routes of entry from Beringia. Several investigators (Fladmark, 1979, 1983; Dixon, 1993; Gruhn, 1994; Erlandson, 2002) have suggested that the earliest human migration to the Americas may have occurred along the north-west coast through the use of watercraft to navigate the oceanic edges of the continent. Coastal stretches may have been deglaciated by about 16,000 B.P. (Mann & Peteet, 1994; Mandryk et al., 2001). The coastal entry hypothesis has recently gained increased support particularly as a consequence of (1) geological research in Alaska and the Yukon (e.g., Jackson et al., 1997; Mandryk et al., 2001), which has established new chronologies for major climatic events in the region and further questioned the existence of the corridor model, (2) the more general acceptance of a pre-Clovis culture, and (3) the discovery of several early coastal sites along the north-west coast of North America and the Pacific coast of South America (Keefer et al., 1998; Sandweiss et al., 1998; Dixon, 2006: Dillehay et al., 2008).
The Tripartite Model
In the 1980s, Greenberg et al. (1986) and Turner (1987) employed linguistic, dental, and genetic data to suggest that the Americas were peopled by three different migrations: (1) the early Amerindians who first reached the interior of Alaska and the Yukon and then spread to some territories of Canada and eventually to the rest of the New World; (2) subsequently, the north-west Pacific coast was colonized by Na-Dene speakers; and (3) lastly, the Arctic was occupied by Eskimos. Although this hypothesis received considerable attention in attempting to explain the peopling of the Americas from an interdisciplinary perspective, it was later strongly challenged on linguistic, bioanthropological and genetic grounds (Morell, 1990; Neves & Pucciarelli, 1991; Szathmáry, 1993a,b; Lahr, 1995; Merriwether et al., 1995) and never had a widespread and lasting impact on the way researchers considered the peopling of South America.
The Dual Migration Hypothesis
During the 1990s, North American patterns of cranial variability were examined and found to exhibit heterogeneous morphological trends, supporting the view that different populations entered the New World at various time intervals. Furthermore, although the closest morphological affinities were with Asian populations, the data could not be used to specify the time of entry or the number of founding populations (Steele & Powell, 1992; Jantz & Owsley, 2001). Recently, a group of South American researchers (Neves & Pucciarelli, 1990, 1991; Neves et al., 2003; Neves & Hubbe, 2005) have suggested, on the basis of a comparison of skulls from Lagoa Santa, Brazil, with late prehistoric and modern Amerindian crania, that craniometric variation in the New World reveals two different, discrete patterns, termed Paleoamerican and Amerindian. Whereas Paleoamericans tend to morphologically resemble present-day Australians, Melanesians, and Sub-Saharan Africans, late prehistoric and modern Amerinds show a greater resemblance to northern Asians. The proponents of this hypothesis claimed that two successive migrations, stemming from different geographical and chronological sources, were responsible for the observed patterns of variation. According to this model, similarities among Australians, Late Pleistocene Asians (e.g., the Zhoukoudien Upper Cave, Minatogawa and Liujiang skulls) and Paleoamericans can be explained by descent from South-eastern Asia (Neves et al., 2003). Paleoamericans are believed to have arrived following a terrestrial route by 15,000 B.P., whereas Amerindians came about 4000 years later along a similar route (Hubbe et al., 2003; Pucciarelli, 2004).
Since the Lagoa Santa skulls have gained a protagonist position in South American physical anthropology, it is useful to review their history in more detail. The crania were initially collected by the Danish explorer Peter Lund in the caves of Lagoa Santa, Brazil, between 1835 and 1844. Lund, based on the Paleolithic chronology of Europe, assigned a great antiquity to the skulls; this was later challenged by McCown (1963) and others. Subsequently, inspired by the putative antiquity of the long and narrow headed Lagoa Santa crania, the French anthropologist Paul Rivet published the morphometrical analysis of 17 skulls of the same “type” selected from a series of 78 undeformed specimens of indeterminate age collected from several rock shelters in Paltacalo, southern Ecuador (McCown, 1963). Shortly thereafter, Rivet published a revision of the “racial” position of a series of modern Pericú skulls from the Baja California Peninsula (Rivet, 1909). The Baja skulls were suggested to belong, together with the Paltacalo skulls, to different branches of a Lagoa Santa “race”, and to resemble Australian and Melanesian crania. (We note, parenthetically, that the same modern Pericú skulls were recently used as evidence for the existence of a “Paleoamerican pocket” in the lower part of Baja California, which apparently survived in the area due to hypothetical climatic changes and geographic isolation during the middle Holocene (González-José et al., 2003).) The Paltacalo skulls eventually acquired a “fuzzy aura of antiquity” among scholars and became, along with the Lagoa Santa crania, the homotypes of the long and narrow-headed ancient Lagoa Santa or Paleoamerican “race” (c.f., McCown, 1963). McCown described in detail further developments of what he called “a curious intellectual edifice”, and concluded that “the initial demonstration of Rivet's race of Lagoa Santa is based upon material that a properly trained modern anthropologist cannot but regard as utterly inadequate” (McCown, 1963).
In order to cope with the problems derived from the disputed antiquity of the Lagoa Santa skulls, for which only two radiocarbon dates have been reported (9720 ± 128 and 9028 ± 120 B.P. non-calibrated; Neves & Hubbe, 2005), Neves generated a more reliable chronological framework of the skulls in the 1990s. Two skulls of the studied samples now date back to ca. 11,000 B.P., namely the Confins skull from Lapa Mortuária (cave) (Walter et al., 1937) and “Luzia” from Lapa Vermelha IV (Prous & Fogaça, 1999). Both skulls were reported to have been dated on charcoal, associated either stratigraphically or directly with the burials (Neves & Hubbe, 2005). Since the first studies were based on a very small sample of crania, additional efforts to increase sample size were subsequently made, assembling a collection of 81 specimens from Central Brazil. The vast majority of the skulls (n = 74) were found to date between 8000 and 8500 years B.P. (Neves & Hubbe, 2005). An exploratory statistical comparison with Howells' database (Howells, 1996) established that the Lagoa Santa skulls resemble present-day Australian/ Melanesian and Africans confirming the results reported at the beginning of the last century by Rivet for his Lagoa Santa “race”. Rivet (1943) interpreted this similarity as proof of an ancient migration to America across the Pacific Ocean, whereas the proponents of the dual migrational hypothesis have postulated a terrestrial route from Africa to America via Asia (Neves & Hubbe, 2005).
In order to be phylogenetically informative, cranial variation should be selectively neutral and not strongly affected by stochastic microevolutionary forces (drift). A number of recent studies indicate that in modern human populations skull size and shape are not predominantly the result of selection, but mainly result from the interaction between gene flow and drift and, to a lesser extent, mutation (Relethford, 2002; Ackermann & Cheverud, 2004; González-José et al., 2008). Operating under the assumptions of a migration/founder model, Powell & Neves (1999) examined a series of Paleoindian and Archaic crania from North and South America and compared these data to a large world-wide sample of late Holocene remains to assess within- and among-group variation and to determine how the observed pattern could be interpreted in terms of population history and structure. Powell & Neves (1999) suggested that their cranial data were consistent with the first Americans having been derived from an undifferentiated Asian population that was not ancestral to modern American Indians. No doubt, this result can be accommodated to support multiple migration scenarios. Nevertheless, according to the same authors, the assumptions underlying the model were not very realistic and when the data were analyzed controlling for the effects of drift, the first Americans were no longer differentiated from modern Amerindians.
Classical Genetic Data
Turning to information provided by synthetic gene-geography graphs based on blood group and protein data, we note that maps of South America based on a small number of polymorphic loci are, as expected, not very informative (O'Rourke & Suarez, 1985). Increasing the number of loci results in maps where the first principal component (roughly 50% of gene frequency variation) reveals clines extending from north-western to south-eastern South America that can be interpreted as indicative of very early population displacements from the Isthmus of Panama to the Brazilian coast and down to the southern cone of the continent (Rothhammer & Silva, 1992; Rothhammer et al., 1997). Maps based on the second and third principal component exhibit clines which are more difficult to interpret, but which may indicate population movements during the Late Archaic/Early Formative along available waterways in search of new alluvial land in the Amazon flood plain, as claimed by Lathrap (1970). Cavalli-Sforza et al. (1994), suggested an early southward migration along the western side of the Andes, consistent with the interpretation that modern speakers of Andean languages may represent descendants of the first occupiers of the region. Lastly, an analysis of 31 protein systems obtained in 29 Amazonian populations agreed with a previously observed differentiation between northern and southern Brazil, suggesting that the Amazon river could have constituted a barrier to north-south gene flow or that genetically differentiated groups may have entered the region from the west (Callegari-Jacques et al., 1994)
Molecular Genetic Evidence
Mitochondrial and Y-chromosome variation in North and South America indicate unambiguously that all Native Americans came from Asia and not from Melanesia, Australia, Africa, or Europe (Karafet et al., 2006; Merriwether, 2006; Wang et al., 2007). Five mtDNA (A; B; C; D and X) and two Y-chromosome (C and Q) founding haplogroups, which are all present among native populations of Siberia account for the molecular variability observed among extant Native American populations (Derenko et al., 2001; Zegura et al., 2004; Starikovskaya et al., 2005). The complete sequencing of 84 mtDNAs have filled gaps in the internal sequence variation of A,C and D, contributing to improve the resolution of the mtDNA phylogeny in Siberia-Beringia (Volodko et al., 2008) Haplogroup X includes many lineages, some of which are present in Europe and Asia. Nevertheless, the lineage found in American Indians is different from those observed in Eurasia (Merriwether, 2006). Ancient mtDNA analysis has until now yielded the same founding haplogroups described in extant populations, thus confirming the genetic continuity between extinct and extant Native American populations (Stone & Stoneking, 1998; Kemp et al., 2007). Consequently, many migration models based on the presence or absence of certain haplogroups have lost much explanatory usefulness (c.f., Goebel et al., 2008).
In an extensive study on patterns of genetic variation in Native Americans from both North and South America, Wang et al. (2007) reported the results of the analysis of 678 microsatellite markers genotyped in 422 individuals representing 24 Native American populations. The analysis of the data revealed a progressive reduction of variability starting from Africa through Asia and down from the Bering Strait to North and South America, consistent with a serial founding African-origin model of human evolution (Prugnolle et al., 2005; Ramachandran et al., 2005). Noteworthy was the observation in South America of a west-to-east difference in genetic diversity, showing that eastern (Brazilian) populations had the lowest levels of heterozygosity. This pattern was also observed with mitochondrial DNA (Fuselli et al., 2003; Lewis et al., 2004) and Y-chromosome markers (Tarazona-Santos et al., 2001). The fact that Brazil exhibits the lowest levels of variation suggests an initial colonization of western South America and a subsequent peopling of the eastern part by western subgroups, confirming the interpretation of the results of a previous genetic analysis of Amazonian tribes (Callegari-Jacques et al., 1994).
Colonization routes also were examined by Wang et al. (2007), by evaluating the relationship between heterozygosity and “effective” geographic distances. When coastlines were treated as preferred routes, there was an increase in negative correlation between heterozygosity and distance from the Bering Strait. A relative genetic similarity between Andean and Mesoamerican populations was also observed by Wang et al. (2007) and was interpreted as consistent with an early coastal colonization. An alternative interpretation is that the two most developed pre-Columbian cultures in the Americas flourished precisely in Mexico and Peru, and that the observed genetic affinities could be the result of late pre-Hispanic (or even early Hispanic) contact between them. We note in passing, that anecdotal information reported by the Spanish chroniclers suggests that in Peru, under the rule of Tupac Inca Yupanqui, long transpacific voyages along the Pacific coast and out to sea were organized using large balsas (rafts) (Rivet, 1943; Cabello-Valboa, 1951; Pietschmann, 1906).
Archaeological Data and Biogeography of South America
The late Pleistocene archaeological record of South America differs from that of North America in many ways (Bryan, 1973, 1986; Dillehay, 1999; Miotti et al., 2003; Borrero, 2006). Among these, there is an absence in South America of a continent-wide stone tool style like Clovis and of evidence for the long-distance movement of raw lithic material. The presence of sharply contrasting environments in the southern hemisphere favoured the establishment of distinct regional cultural traditions at the outset of human dispersion. Furthermore, the extensive ice sheets covering high latitudes limited the movement of colonizers in North America in contrast with South America where the glacial effect was confined to patchy high altitude and latitude areas (Clapperton, 1993). The archaeological evidence of human entry into South America before 15,000 B.P. is weak and only presumed at this time. However, given the evidence confirming the presence of humans probably centuries before 12,500 B.P., the likely entry date should lie between 15,000 and 13,500 B.P.
Between at least 11,000 and 10,000 B.P. South America witnessed several major changes, including the increased use of coastal resources, demographic concentration in major river basins, and modification of landscapes and plant and animal distributions. Some 1000 years later an increase of site density, technological diversification, plant domestication, and appearance of ritual practices indicate that the initial impulses toward cultural complexity had developed in South America, probably within only a few millennia after the initial arrival of colonizers (Dillehay, 2000; Lavallee, 2000). No doubt, sharp contrasting environments characterized the continent during the late Pleistocene and Pleistocene to Holocene transition and contributed to cultural developments that show a steady shift away from sub-continental uniformity and toward the establishment of distinct regional, cultural, and especially technological traditions (Bryan, 1973, 1986; Prieto, 1996; Dillehay, 1999; Borrero, 2006). The impact of climatic and environmental changes on these developments is not well understood. There is evidence of a significant temperature rise between 15,000 and 14,000 B.P. and a subsequent warming trend lasting until 10,500 B.P. (Rull, 1996; Prieto, 1996; Latrubesse & Rambonell, 1994; Weng et al., 2004; Bush & John, 2006).
South America is biogeographically divisible into four major areas: the Andes, the tropical and temperate plains of Colombia, Venezuela, and Brazil, the highlands of east Brazil and the Guyanas, and the southern Pampa and Patagonian grasslands. These divisions are important for considering hypothetical migration routes into the southern hemisphere (c.f., Bennett & Bird, 1964). The Andean mountain chain, the vast Amazonian basin, and the southern grasslands were doubtless a combined barrier for isolating some human populations geographically and genetically, regularly disrupting gene flow between geographically contiguous populations, and producing some of the regional skeletal and genetic differences discussed earlier. Although foraging mobility undoubtedly continued throughout the terminal Pleistocene to early Holocene periods, reduced mobility in some regions probably derived from the increasing use of local resources constitutes one potential explanation for the bioanthropological heterogeneity observed across the continent. Both genetic and morphological signals of marked diversity between the Andes and the eastern lowlands of the continent, generally agree with differences observed in the technological and subsistence patterns in these two regions.
Continental ice sheets started to melt and the sea levels began to rise between 14,000 and 13,000 B.P. The Pacific and Atlantic shelves and many areas in Tierra del Fuego were flooded, as were any archaeological sites, making the study of early migration routes along the coasts very difficult. These and other late Pleistocene environmental changes introduced significant resource-related changes that likely favoured social learning and bonding and technological change, as evidenced by a wide array of local and regional stone tool technologies and subsistence patterns in place by at least 10,500 years ago (Kipnis, 1998; Dillehay, 2000; Salemme & Miotti, 2002; Flegenheimer et al., 2006; León-Canales, 2007). The broad network of environmental responses by people across the continent clearly included more than the development of social bonding and strategies for sharing resources. There is so much diversity across the continent at this time and later that it is difficult to say to what degree the eastern lowlands were culturally different from the Andean region or the southern grasslands were from the Amazon basin or northern tropical forests. Regional technological diversity suggests that there were specific responses by different groups to the great variations in local habitats and raw material availability that existed across the continent, especially in the Andean region. Technology offered a flexible, rapid means of responding to climate and other changes at a range of scales from daily needs of individuals to long-term trends experienced over generations.
Several sub-continental studies have emphasized the instability of climates during terminal glacial, deglaciation, and the transition to the early Holocene from ca. 13,000 to 10,000 years ago (Bennett et al., 2000; Bush & John, 2006; Lovell & Kelly, 2008). For humans, these cyclical periods of climatic stability and instability would have placed additional selective pressures on integrated biological and behavioural responses to increased resource variability. At the continental scale, populations distributed across Amazonia or along the Pacific coastal shelf during warm deglaciation, possibly would have become geographically isolated during times of augmented aridity. Increased precipitation and vegetation growth in some areas after 11,000 B.P. may have pushed some populations outward towards the better vegetated and watered areas, such as the Amazonian rainforest and the mid-latitudes and western altitudes of the continent. Such repeated cycles and disruptions of landscapes were doubtless a force for isolating populations geographically, biologically, and consequently genetically, regularly disrupting gene flow even between contiguous populations and more so between those on the east and west sides of the continent. At sub-continental scales, expansions and contractions of environments during the terminal Pleistocene and early Holocene period (ca. 12,000–10,000 B.P.) would have drawn in or pushed out human populations from north to south, or vice versa, as well as eastwards into the Amazon basin during times of expansion or retreat from aridity, for example. The disruption of contact between regional populations combined with occasional severe downturns, causing local animal and possibly human extinctions, also would have contributed to the development of regional differences in skeletal morphology, gene pools, and social and technological behaviours. Although social processes must also be taken into consideration to explain these differences in the morphological, genetic, and archaeological records, for which there is currently no hard evidence, the periodicity and extremes of late Pleistocene climatic cycles must have underlain a process of accretionary or mosaic emergence of human and cultural variation within South America.
Considering the various sources of evidence discussed above one can outline a possible scheme of human dispersion in South America. Hunters and gatherers who migrated to South America via the Isthmus of Panama could have entered the Andean highlands by way of the Cauca and Magdalena river valleys which flow from south to north in Colombia. Once adjusted to the highland environment there would have been few barriers for moving further south, although some high altitude areas were covered by glaciers. Some groups may also have migrated eastward following the Caribbean rim of Venezuela, the Guyanas, and north-east Brazil and others into the interior of Venezuela and afterwards south-east or south-west along several large river systems into the Amazon basin. People may also have migrated along the Pacific coast down to Chile, following favourable fishing localities and using watercraft (Rothhammer & Silva, 1992; Cavalli-Sforza et al., 1994; Keefer et al., 1998; Sandweiss et al., 1998; Stothert, 1998; Wiesner, 1999; Wang et al., 2007). From the Amazon basin and/or the Andes of north-west Argentina, people could have entered the open parkland country of eastern Brazil and also could have spread throughout the Pampas and Patagonia. As mentioned above, genetic data from Native Brazilian populations suggests that the Amazon River constituted a barrier to north-south gene flow. There is also the possibility that genetically divergent groups may have entered the region from Colombia, Venezuela, and the Guyanas in the north and from Argentina in the south (Fig. 1).
There is evidence in parts of north-western and central South America that several plant species were deliberately manipulated and possibly domesticated in terminal Pleistocene times (Piperno & Pearsall, 1998; Piperno & Stothert, 2003). What is not understood is whether there are any connections between broad-spectrum diets in environments where potentially domesticable plants occurred, specific resource zones, and social changes that might have favoured the first impulses toward food production. Similar connections need to be ascertained for the kinds of human and camelid interactions that took place during this period in the puna and altiplano zones of the central and south-central Andes. Diets, resource zones, and local human population histories constitute only some of the variables required to understand the development of early cultural diversity, food production within a relatively short period of time after humans arrived, in comparison to the Old World, and how ecological constraints and opportunities played out in setting the stage for subsequent sedentary and later more complex societies in South America.