Among the theories concerning the peopling of the Americas, the ‘Three’ [14] and the ‘Four’ [25] and [26] migration models (TMM and FMM, respectively) are the mostly accepted. In both models, the term ‘Amerindians’ refers to all American Indians, with the exception of Na-Dene, Eskimos and, according to the FMM, Paleoamericans – the most ancient inhabitants of Americas. They propose that Amerindians are descendants of a unique migratory wave coming from the North-East of Asia, so they would constitute a homogeneous population. Nevertheless, some genetic [3], [9], [28] and [35], linguistic [27] and morphologic [2], [17], [21] and [34] studies suggest that Amerindians do not behave as predicted by these models. The aim of this study is to assess the homogeneity of Amerindian craniofacial morphology in a worldwide context. According to the TMM and the FMM, it is expected to find low levels of morphologic variation among southern Amerindians.

Assessment of morphologic variation was based on a “neutral model bound” approach [31] and [32]. According with Relethford [30], there are two ways to assess the regional variance: the geographic regions or the local populations of each region can be taken as the unit of analysis (among-regions and among-local populations analyses, respectively). In the first case, a unique F _ST for the entire comparative context is obtained which measures the proportion of the total variation that is due to differences among the units analysed (regions). The variation is deduced from the observed variances. In the second case, a F _ST for each region is obtained. It has the advantage over the first case that it allows to deduce which part of the entire variation in a geographic region is due to the among-local populations differentiation.

For the among-regions analysis, seven world regions (South-Saharan Africa, Europe, Australasia, Asia, Polynesia, Americas, and Patagonia) each one represented by three local populations were compared. For the first six regions, data come from the Howells’ database and the selected populations (Table 1) are those used by Relethford [30]. For the seventh one, Patagonia, measurements were obtained by one of us (MLS). Patagonia seems to have been occupied by a unique migratory wave. Even when the contact with northern groups has been demonstrated [12], an intense gene flow was improbable and moreover, this region never underwent an important demographic growth [1]. Thus, a low variation is expected among Patagonians. They are represented by two northern samples: Chubut River Valley (CH, n = 99) and Rio Negro River Valley (RN, n = 72), and a southern one: Tierra del Fuego (TF, n = 45) (Fig. 1). The among-local populations analysis was made with the samples included in the among-regions analysis. A second among-local populations analysis was carried out with Peru, CH, RN and TF and with three additional samples (Fig. 1): the Bolivian sample (BO, n = 19) of ‘aymará-quechua’ origin, localised in the Andean highlands; the hunter-gatherers of the Paraná River Delta (PD, n = 38), with an antiquity estimated by archaeological associations around 2000 years BP [39]; and the western Pampas sample (WP, n = 37) which represents a Mapuche population that settled in the Pampas after migrate from the southern Andean region [4]. Data for these three samples were collected by one of us (MLS).

Forty-three craniometrical linear Howells’ variables [18] were used in the study (Table 2). They were transformed in dimensionless vectors through the Q-standardisation [5] to remove size, whose variation among Amerindians is attributed to environmental factors [17], [21] and [34]. Following previous studies [31] and [32] additive genetic variance-covariance matrixes, from the phenotypic ones, were calculated with a heritability of 0.55. This value was estimated by Devor [7] and represents an average proportion for metric characters. In the among-regions analysis, the regional variation was calculated through observed variances; in the among-local populations analysis, the regional variation is represented by the F _ST values.

In the among-regions analysis, an unbiased F _ST of 0.1492 (standard error = 0.0015) and a mean variance of 0.819 between all the regions were obtained. Patagonia and South Saharan Africa show the greatest observed variances (Table 3). In the among-local populations analysis, the F _ST values are very similar to those obtained by Relethford [30] who found a significant association of the among-populations differentiation and the geographic distances. In this context, Patagonia occupies the third position after Polynesia and the Americas showing a higher variation than larger regions (Table 3).

F _ST values were obtained with BO, WP, PD, CH, RN, TF and the Howells’ Peru samples. Each combination of three South American samples produced thirty-five normally distributed F _ST values. The minimum value is that of Patagonia (0.0890), the maximum 0.2180 represents the comparison of WP, PD and Peru. The mean of the distribution is 0.1573 with a standard deviation of 0.0323. F _ST values for the geographically closest groups are: 0.135 (CH, RN, PD), 0.162 (CH, RN, WP), 0.179 (RN, PD, WP) and 0.203 (CH, PD, WP).

Results indicate that the South Amerindians are far from being a homogeneous group in cranial morphologic terms (Table 3). Gene flow with Europeans should be discarded due to the absence of morphologic affinities between Europeans and these Amerindian groups [34]. The low variation among Patagonians regarding other Amerindians supports their unique origin inside America. However, it is worth to know that these same groups from Patagonia show the first and the third highest variations in a worldwide context (Table 3). The high variation in Amerindians can be associated to: (a) a greater antiquity of the peopling, (b) several waves of migration, (c) greater effective population size, (d) greater rate of population growth, (e) greater degree of isolation, and (f) adaptative factors.

The option a), a peopling before 15000 years ago, has been suggested by many authors [15], [16], [27] and [33]. This option seems probable even when archaeological sites older than 13000 BP [16] are not accepted without doubts [8].

Option b) is difficult to accept in the context of the TMM and FMM models that suggest that Amerindian characteristics must reflect the northeastern Asiatic ones. However, in Siberia or Beringia a high variation is not reflected before Holocene times. Kozintsev et al. [20] mentioned that Baikalians, about 8000 years BP, had extremely flat faces and the later inhabitants were less “mongoloids” due to gene flow with European populations. But for both the TMM and FMM the arrival of Amerindians was at the Terminal Pleistocene, so the Siberian holocenic diversity does not explain, therefore, the Amerindian one.

If another than an eastern Asiatic origin is considered, the morphologic diversity is easier to be explained. Genetic studies support the Asiatic origin of American Indians [6], [23] and [24] without accord about the geographic localization of the ancestor [11], [22], [23] and [38]. However, some Amerindian tribes show genetic characters that do not accommodate with the suppositions of “homogeneity”. Brown et al. [3] found a haplogroup X extensively distributed in America and present in Asia only in the Altaians of South-Central Siberia [10]. Even when there is not conclusive evidence that Amerindians have more than one ancestor – non-Asians – the Amerindian diversity does not seem to be completely derived from the Asiatic one.

By other hand, if the double-migratory event (Paleoamerican and Amerindian) for the main part of the Americas is accepted, the eventual genetic relationships between both waves must be discussed. Morphologic divergence between Paleoamericans and Amerindians would suggest that the former have not originated or at least did not genetically contributed to the later [19], [25], [26] and [36]. However, Powell and Neves [29] proposed that micro evolutionary processes could be responsible for that morphologic divergence. It is also probable that some South American groups are direct descendants of Paleoamericans as it was suggested for the Pericúes of Baja California in Mexico [13].

Options (c), (d) and (e) are proposed by the genetic neutral models. Options (c) and (d) do not seem probable: according with Rogers et al. [33], the hunter-gatherers increase very rarely their population size in a short period of time moreover during climatic instability. Steele et al. [37] developed a model for Paleoamerican dispersion in North America during the Pleistocene-Holocene transition and they deduced that it was not followed by an increase of the population size.

Option (e) is partially related with population size. The archaeological record shows that around 13000 BP South America was widely occupied. Due to the arid conditions after 13000 BP, populations would remain near waterways sites limiting their mobility [8]. Thus, the fragmentation of the geographic range would contribute to variation through genetic drift.

The option (f) was exceptionally mentioned [17] and [21] and it is discarded. The effect of environmental factors on morphologic traits is not denied but its contribution to inflate or reduce the distances among populations has not been proved. Moreover, it is difficult to understand why the environmental pressure would be greater in the Americas than in any other region.

These results show a high level of craniometric diversity among the southern Amerindians. This fact would result by different but not exclusive processes: (1) a low population size of first migrants followed by a great dispersion in different and unstable environments; (2) the genetic contribution of the ancient Paleoamericans to the most modern Amerindians; (3) many migratory waves coming from different geographic regions; and (4) an old settlement (older than 13 000 yr) of Americas. The high craniometrical variation found in this study is insufficient to solve the problem of their evolution but it constitutes an element that questions the established idea about Amerindians.