The Southern Divide : Testing morphological differences among bifacial points from southern and southeastern Brazil using geometric morphometrics

Bifacial points have been used to characterize the “Umbu” tradition in southern and southeastern Brazil. This archaeological tradition has been related to sites dated from the late Pleistocene-early Holocene boundary to near historical times. Such a huge temporal range and vast territory have suggested the existence of greater diversity within this tradition that has been ignored thus far due to the lack of systematic regional studies of such points. Through geometric morphometric analysis, this article aims to test the hypothesis that there are substantial differences in the Holocene bifacial points associated with the Umbu tradition in southeastern Brazil. Five landmarks were digitized in standardized photographs from 658 points from the states of São Paulo, Paraná, Santa Catarina, and Rio Grande do Sul. The results show that points made by groups from southeastern Brazil (São Paulo state) present a very distinct morphology (size and shape) in comparison to those made by the southern groups (Paraná, Santa Catarina and Rio Grande do Sul). This would indicate a regional identity shared only by some groups from São Paulo (at least regarding the projectile points). It is possible that Umbu tradition presents a more restricted range, both in chronological and spatial terms, than the one proposed so far.


Introduction
Flaked lithic artifacts shaped as projectile points are quite common in the Americas, being generically called "arrowheads" or "projectile points".Although their piercing or puncturing function is obvious, there are different ways in which such piercing or drilling can be carried out (Cattelain 1997): pressure-induced muscle strength of those handling the artifact directly transmitted via a shaft (thrusting spear), pressure-induced shock between the artifact and the target, being the kinetic energy released by either a bow or a dart thrower.It is known that stone artifacts that look like points can also be used as awls, knives, spears, harpoons, among other uses (Pitt Rivers 1906: 101, 117;Fenenga 1953;Rausing 1967: 164;Kay 1996;Greaves 1997;Erlandson et al. 2014).In short, the terms "arrowhead" or "projectile point" are inadequate, since they presuppose the use of a specific device to transmit the energy (For a recent review on the potential propulsion systems associated to bifacial points from southern and southeastern Brazil, see Okumura 2015 andOkumura &Araujo 2015a).Therefore, in this article, we will designate such artifacts as "bifacial points".An important feature of bifacial points lies in the fact they are formal artifacts, i.e. instruments whose manufacture aims to obtain a specific patterned form.
In the 1970s, Brazilian archaeologists tried to understand the morphological diversity observed in bifacial points through the use of typologies and the creation of archaeological phases (see Table 1).Miller (1967;1974), working in the Sinos Valley and Maquiné region (Rio Grande do Sul), created three pre-ceramic phases that presented bifacial points: Camuri, Umbu, and Itapuí.The Camuri Phase was characterized by open-air sites, while Umbu and Itapuí Phases were associated with rockshelters.Umbu would be more ancient (6000 to 4000 years BP) than Itapuí, and the predominant bifacial point type would be stemmed points with triangular bodies, as well as lanceolated points.Itapuí Phase would be more recent than Umbu (4000 to 1000 years BP), being characterized by points with bifurcated stem and triangular body, sometimes presenting serrated edges.
Another attempt to classify sites presenting bifacial points was the creation of the Bituruna tradition (Chmyz 1981a), which was tentatively associated to Paleoindian sites, where large stemmed bifacial leaf-shaped points were abundant.This tradition was identified in the middle and low Iguaçu River, in central-southern and south-western Paraná state, as well as in the areas near powerplants in Foz do Areia, Salto Santiago and Salto Caxias (Chmyz 1969;1981a;b;Parellada 1999).However, the alleged antiquity of this tradition still remains to be verified; thus far, dates are scarce.The only radiocarbon age concerning this tradition comes from the site Jusante UHE Salto Caxias I, in south-western Paraná, presenting a date of 4810 ± 360 BP (Australian National University-ANU 192-19;Parellada 2005: 30).
Many other archaeological phases presenting bifacial points have been identified in southern Brazil (Iguaçu and Potinga Phases;Chmyz 1969;Vinitu Phase;Kern 1981: 215-220;Schmitz 1984: 12-14;1991a;Itaguajé Phase;Chmyz & Chmyz 1986;Itaió Phase;Piazza 1974;Capivara Phase;Schmitz 1991a).Despite these early efforts to sort out the morphological diversity observed in the bifacial points from southern Brazil, the lack of good chronologies and the unclear definitions for the bifacial point classes (sensu Dunnell 1971: 45) led to difficulties in assigning new sites to these "phases", resulting in a later lumping of them into a single "tradition", Umbu.
The Umbu tradition would present its oldest phase in southwestern Rio Grande do Sul with the Uruguai Phase, in the 11 th Millenium BP (Kern 1981: 232-8;Schmitz 1987) (see Table 1).A more recent revision of the sites and dates by Dias & Jacobus (2003) considers only ten sites with a chronology ranging from 10,800 to 8500 BP.If we accept the revised data, Uruguai Phase would be contemporaneous to Umbu, since the oldest date for Uruguai Phase is 10,985 ± 100 BP (Laranjito Site, see Table 1).According to Schmitz (1978: 108), the Uruguai Phase would be the beginning of a tradition presenting bifacial points that would continue until the 14 th Century with the name Umbu (while Schmitz (1978: 108) supports the idea that Uruguai Phase would represent the beginnings of Umbu tradition, other authors, like Hilbert (1991: 16), consider that Uruguai Phase would be followed by Umbu tradition.Schmitz (1978:112) also notes the similarity among bifacial points from Uruguay, Missiones (Argentina) and Southern Brazil.Schmitz (1987) draws attention to the fact that, despite the large number of sites identified as representatives of Umbu tradition very few chronological indicators have been identified so far.
This eventually generated controversy concerning the existence of such a tradition, since it would encompass a very extensive chronological range, from 11,000 to 500 years BP.Therefore, Umbu tradition poses a special problem to archaeologists because of two basic characteristics: its wide geographic distribution and, above all, the length of its chronology.Researchers have questioned whether identification of a tradition extending from the Pleistocene-Holocene period until the eve of the historical period is plausible.Such doubts have led to the discussion of the validity of the concept of "Umbu tradition" in terms of classification and organization.Given the lack of identifiable chronological markers, this tradition became, in practice, "defined" only by the presence of bifacial points, becoming synonymous with "sites presenting bifacial points", regardless of their chronology or geographic location.
Currently, this tradition, dispersed throughout southern and southeastern Brazil, is simply characterized by the presence of bifacial points, presenting dates ranging from the Late Pleistocene to historical times (Schmitz et al. 1980;Schmitz 1999;Noelli 2000).Therefore, throughout the years, all archaeological sites presenting bifacial points, from Rio Grande do Sul to São Paulo (and in some cases including more northern settings, like Minas Gerais, (Koole 2007;2014) and Mato Grosso do Sul, (Kashimoto & Martins 2009;Martins & Kashimoto 2012)), ended up being classified as belonging to the Umbu tradition.For example, in the case of São Paulo state , Miller Jr. (1972) defined the Rio Claro tradition (based on sites from Rio Claro, São Paulo state) as presenting bifacial points in some phases.Schmitz (1978: 120;1991) also remarked the differences between the material from Rio Claro and Uruguai Phase, considering the bifacial points from Rio Claro as a regional evolution of Umbu tradition.Later, Prous (1991: 154) reports that this tradition would have been subsumed within Umbu tradition.This inclusion appears to be quite controversial since not all phases of the Rio Claro tradition presented bifacial points, and thus theoretically it could not be included in Umbu tradition.Regardless of the adequacy of this tradition, it is likely that such direct association have been obliterating important regional and local variations, either in chronological or spatial terms.To this date, regardless of the name given by archaeologists, it is possible to say there is a substantial number of archaeological sites in southern and southeast Brazil presenting bifacial points dating from the Late Pleistocene to the eve of the historical period (Table 1 and Figure 1).Figure 1: Archaeological sites from southern and southeastern Brazil presenting bifacial points dated from the Late Pleistocene to the eve of the historical period.The numbers refer to the column "map number" in Table 1.

Classification and the problems of assigning archaeological materials into "archaeological traditions"
The origins of the concepts of "phase" and "tradition" can be attributed to the Midwestern Taxonomic Method, by which North American archaeologists in the 1930s (McKern 1939) tried to define (or replace) the concept of "archaeological culture".Decades later, a similar system was proposed by Phillips and Willey (1953), in the 1950s and partially adopted in Brazil by "The National Program of Archaeological Research" (PRONAPA -Programa Nacional de Pesquisas Arqueológicas) in the 1970s.According to PRONAPA (Chmyz & PRONAPA 1976: 145), a tradition is defined as "a group of elements or techniques that are distributed with temporal persistence."In principle, this concept was totally disconnected from any "ethnographic" meaning (similar to that proposed by the Midwestern Taxonomic Method).Archaeological research pioneers aimed at defining traditions, which was often done from the study of only one or two sites.Later, phases and traditions began to be compared to "autonomous and semi-autonomous units" or "tribes" ("phases"), "tribal or linguistic entities" and "nations" ("traditions" - Meggers & Evans 1985;Schmitz 1991b).
In Brazil, criticisms of the definition and use of the term "tradition" pointed out three major shortcomings (Dias 2003: 51;Dias 2007;Dias & Hoeltz 2010;Hilbert 1994;Milder 1999): 1) a definition of such traditions was based on a few typological criteria, 2) the use of a fossil guide to determine the association of a particular site to a tradition and 3) the use of few attributes for classification of a site in a tradition.In other words, the application of the term "tradition" resulted, for example, in all sites presenting bifacial points (which are the fossil guides of Umbu tradition) being automatically classified as belonging to the Umbu tradition, regardless of the morphology, chronology, geographic location, or type of site.
After more than four decades since the initial definitions were proposed by PRONAPA, there were subsequent criticisms, followed by the remodeling of the "Pronapian" concepts, seeking stronger ties with anthropology, followed again by new waves of criticism, and finally today we have the crystallization of the concept of "tradition", simply because there is nothing better to replace it.After 40 years of controversy, perhaps we can be at peace with the term "tradition", as long as we understand it as a heuristic tool.Indeed, perhaps the best definition of tradition is the original one: simply a "group of elements or techniques that are distributed with temporal persistence".Fortunately, such definition refers to a group (a cluster of elements that can be listed) rather than a class, which would require a definition (Dunnell 1971: 45).Thus, the traditions will never be defined, only described.This may be unsatisfactory from a formal point of view, but again, perfectly serves our heuristic purposes.
Therefore, we do not propose abolishing the use of the term "Umbu tradition".However, we want to emphasize the importance of recognizing a morphological diversity present in these bifacial points that has been ignored so far.Traditions, or whatever name we give to these aggregate of phenomena, can be useful in terms of transmission of information among professionals and even among the laymen.However, they can be harmful (not to mention useless), when they ignore the observed variation by simply lumping together different classes of artifacts.In this context, we believe that the (ab)use of the term "Umbu tradition" must undergo further reflection.Binford (1977;1979) suggested that technological organization could be seen as a continuum between cases centered on the production of highly modified ("curated") tools and cases where the tools are made from slightly modified raw materials ("expedient").However, a quick review of the literature clearly points to a greater emphasis on characterization and study of formal lithic industries, whether in the Old World or the New World.Certainly, there is still a fascination for the formal lithic industries, reminiscent of collecting and curio cabinets, since these artifacts are more visually and aesthetically appealing, besides being more easily recognized as proper artifacts.Another important point is that, based on the discussions on style and function (Binford 1977;1979;Dunnell 1978;Sackett 1985;among others), formal artifacts are more easily "seized" as conveying cultural transmission processes.This led to the development of various systems of classification, resulting in a greater availability of statistical analysis that can be applied in order to characterize such industries.This is the case in Europe and North America, for example, where the emphasis is on the variation between "types" of formal artifacts, which are generally considered independent of the technology.This emphasis has been the hallmark of the lithic analysis developed by François Bordes (1950), and this perspective has permeated the archaeological thought far beyond its original application in Middle Paleolithic European assemblages.Currently, many archaeologists may not share the fundamental ideas of Bordes, which sought to determine ethnicity and social interaction between different cultures, but still shape and technique are considered independent entities (Draper 1985).Thus, one cannot ignore the explanatory potential of formal artifacts, since this characteristic, the standardization of gestures and techniques aimed at producing artifacts with specific forms, allows the tracing of cultural interactions.

Why study bifacial points? Formal artifacts and their potential statistical approaches
According to Dias (2003: 225), a lithic industry can only be fully understood with the analysis of the entire operational chain, and the typological variation observed in formal artifacts is just the tip of the iceberg (Perlés 1992: 223-224).However, what is proposed in this article is not an in-depth study of lithic technology itself, but the presentation of a complementary approach, involving the use of geometric morphometrics and multivariate statistical tests not very often used in Brazilian archaeology (but see Okumura & Araujo 2014), but whose potential has been exploited successfully in several studies abroad (Saragusti et al. 2005;Cardillo 2006;2009;2010;Buchanan et al. 2007;Castiñeira et al. 2009;2011;2012;Franco et al. 2009;Archer & Braun 2010;Buchanan & Collard 2010;Costa 2010;Lycett et al. 2010;Iovita 2011;Brown et al. 2012;Charlin & González-José 2012;Lycett & Von Cramon-Taubadel 2012;Thulman 2012;Wang et al. 2012;de Azevedo et al. 2013;Lycett & Eren 2013;Charlin et al. 2014;Davis et al. 2015;Fox 2015;Cardillo et al. in press).The use of multivariate statistical methods applied to geometric morphometrics data might help elucidate old questions about the characterization of the Umbu tradition, possibly clarifying the meanings related to changes in the morphology of bifacial points over time, the chrono-spatial relationships among different sets of points, among other issues.In this article, we will be focusing on testing the hypothesis that there are important differences in the morphology of Holocene bifacial points associated with the Umbu tradition in southeastern Brazil.Preliminary results regarding this sample have been previously published (Okumura & Araujo 2013).

Geometric morphometrics: An overview
The analysis of the morphology has always played an important role in different areas of knowledge, including biology, arts, and engineering, among others.Differences in morphology can be briefly described through familiar objects such as geometrical shapes or letters of the alphabet.However, such descriptions are rather vague, inaccurate or even erroneous, especially when the shapes are complex (Zelditch et al. 2004: 1) and when an appropriate morphometric approach is not applied.Although approaches using linear measures are often used to characterize the morphology of formal artifacts, it is well known that important information, especially regarding the shape (in comparison to size) is lost (Zelditch et al. 2004: 5).In order to avoid losing information on the form (constituted by shape and size), we conducted an analysis using Geometric Morphometrics (GMM).
GMM "is a disparate set of techniques with a common purpose: the statistical analysis of differences in form using a quantitative description that preserves the geometry of shape variation" (Viscosi & Cardini 2011: 3).This preservation allows the visualization of group and individual differences.In this sense, taking the form, it is possible to separate size from shape, to quantify shape and to test differences among shapes (Bookstein 1991).The data obtained using GMM are coordinates of shape landmarks, whereas the traditional morphometrics deals with distances between landmarks.Therefore, GMM aims to quantify the differences in morphology through the use of landmarks (Bookstein 1991).Landmarks can be defined as "samples of discrete points which correspond among all the forms of a data set" (Rohlf & Bookstein 1990: 63) or "discrete (…) loci that can be recognized as the same loci in all specimens in the study" (Zeldich et al. 2004: 23).The latter definition implies landmarks in a biological context.According to Zeldich et al. (2004: 24), landmarks should not change their topological positions relative to other landmarks (a one-to-one correspondence in the specimens to be compared) (Viscosi & Cardini 2011), should result in a good coverage of the studied morphology, should be observed repeatedly and reliably, and should present coplanarity.The form of the structure is captured using the Cartesian coordinates of a configuration of landmarks.According to Rohlf & Bookstein (1990: 220-221), there are three categories of landmarks (but see Valeri et al. (1998) and Gunz et al. (2005) among other authors for other different types of landmarks).Type 1 landmarks are "discrete juxtapositions of tissue types".This kind of landmark is preferred because there is no need to mention any structures far from the landmark itself.Type 2 landmarks are considered more problematic, because they are identified as "maxima of curvature or other local morphogenetic processes".In archaeology, the majority of the landmarks are Type 2. Type 3 landmarks are described as "extremal points", which renders their use very problematic, mainly because they are taken as endpoints of "as farthest" from other points.The importance of distinguishing the type of landmarks rests not only in the intrinsic quality of each type (Type 1 should be easier to be observed in a repeatedly and reliable way), but also in the amount of information that can be retrieved from each type.Because Type 1 landmarks are located in the middle of different structures, it is possible to identify the directions of forces acting upon that area.The same information cannot be obtained from the analysis of Type 2 landmarks, because they are not completely surrounded by structures (Zeldich et al. 2004: 31).

Geometric morphometrics applied to the analysed sample
In this article, we explore some of the potential applications of GMM in the analysis of the morphology of bifacial points.Five Type 2 landmarks were distributed in order to include the different parts of a bifacial point (Figure 2).The five landmarks satisfactorily cover half of the specimens, which is a frequently used way to reduce the time of data collection in symmetric structures (Viscosi & Cardini 2011).These five landmarks were digitized in standardized taken photographs using the software TPSDig2 (Rohlf 2015).The photographs were taken with the camera parallel to the projectile point surface.Points were laid flat with their distal ends facing to the right and a metric scale was also included.Virtually flat things like bifacial points can be reasonably analysed using a two-dimensional approach without losing much important information (Velhagen & Roth 1997;Buchanan & Collard 2010).The sample included 658 bifacial points from São Paulo, Paraná, Santa Catarina, and Rio Grande do Sul.Photographs were taken from complete and finalized bifacial points, meaning that broken points and preforms were not included in the study.The description of each group is presented in Table 2 and Figure 3 shows the geographic location of groups.Landmarks were later transformed into shape coordinates using Procrustes method.Geometric morphometric analyses were carried out using TPSRegr, TPSSmall, TPSRelw and TPSPLS (Rohlf 2015).For a detailed description of GMM principles, see Okumura & Araujo (2014).Table 2: Geographic origin and number of points included in the analysis.Group names in bold refer to groups composed exclusively of points of a single archaeological site, in this case, it was decided to name the group according to the name of the archaeological site.

Results
In GMM, the centroid size measures the dispersion of landmarks around the centroid of the configuration.Centroid size is a measure of size that is mathematically independent of shape (Zeldich et al. 2004: 13) and it was computed using the five landmarks described in the Materials and Methods section.An ANOVA indicated significant differences among groups (F = 10,081).A post hoc Bonferroni test (Table 3) shows that two groups of São Paulo (Alice Boer and Ipeúna) are similar to each other and significantly different from the others, with the exception of Campos Novos (SC), Ibicuí (RS), and Taió (SC). Figure 4 shows that points from São Paulo are, in general, bigger than those from the south.Figure 5 presents the Principal Component Analysis applied to the shape coordinates.There is a overlap among the groups.In this case, the consensus shape presented in the upper corner would be located in the center of the chart (coordinates 0,0).It is possible to verify that there are points whose body is long and stem is tapered (points from the upper portion of the graph); points whose body is short and the stem is tapered (points from the right side of the graph); points whose body is short and the stem is forked (lower portion of the graph); and points whose body is long and the stem is forked (left portion of the graph).A higher frequency of short bodies observed in some samples possibly indicates episodes of resharpening.The importance of resharpening, which relates to an allometric relation between shape and size is also indicated through the high correlation observed between the aligned data and centroid size (0.50721).
From the Relative Warps Score Matrix, we performed a Canonical Analysis using the entire matrix.The graph representing the two functions of this canonical analysis (Figure 6) shows that the two groups of São Paulo present a different shape compared to the southern groups.It is also possible to verify an association between Paraná and Santa Catarina, especially Reserva (PR) and Taió (SC).The morphological similarity between points from these two regions was also observed in the analysis of linear measurements (Okumura & Araujo 2015b).

Discussion and conclusions
The GMM analysis point to the presence of significant differences in the size and shape of the bifacial points from São Paulo compared to the southern region.The bifacial points from the two groups from São Paulo (Rio Claro and Ipeúna) presented high morphological similarity.In general, the bifacial points from São Paulo seem to be larger than the other groups (with the exception of two groups from the south: Campos Novos and Ibicuí).Such results have been previously observed through the analysis of linear measurements of points from Rio Claro and southern states (Okumura & Araujo 2015b).Although we believe it is not necessary to revive the category "Rio Claro tradition", our preliminary results point to an important difference in the morphology of the points from São Paulo in relation to the southern points of the country.Since Umbu tradition was defined based on the material found in southern sites, the points from São Paulo could not be considered part of this group.It may be possible that the points from Minas Gerais and Mato Grosso do Sul are also distinct from the points of the southern region.
There is evidence of similarity between the points from Paraná and Santa Catarina, as well as between some groups in Rio Grande do Sul (Okumura & Araujo 2013;2015b).Such similarities could be exacerbated due to the huge difference between the points from São

Figure 2 :
Figure 2: Drawing presenting the different parts of a bifacial point and landmark configuration used to characterize the different parts of a bifacial point.(A) the apex of the body in the longitudinal line (distal end), also defined as the junction of the two blade edges, (B) the most extreme point in the shoulder curve, (C) the point where the neck meets the body, (D) the meeting of the lateral and the basal parts of the stem, and (E) the most extreme point of the stem in the longitudinal line.

Figure 3 :
Figure 3: Map presenting the geographic location of the analysed groups.

Figure 4 :
Figure 4: Box-plot graphs based on centroid size.

Figure 5 :
Figure 5: Principal Component Analysis applied to the shape coordinates.Upper corner: consensus shape (note that the program output represents half of the point).

Figure 6 :
Figure 6: Graph representing the two canonical functions.Crossed circles: São Paulo state, parallelogram: Paraná state, circles: Santa state, and squares: Rio Grande do Sul state.The first function explains 53,4% of total variance and the second function explains 35,5%.

Table 1 :
Archaeological sites from southern and southeastern Brazil presenting bifacial points dated from the Late Pleistocene to the eve of the historical period.The column "number on map" refers to the map presented in Figure1.