RESEARCH ARTICLES her collaborators at the first Olmec capital, San Lorenzo, Veracruz (5), emphasize the Olmec Pottery Production and Export higher sociopolitical level that the achieved relative to contemporaneous groups in Ancient Determined in Mesoamerica (Fig. 1). While leaders, for example, at the emerging chiefdom of San Through Elemental Analysis Jose´ Mogote in the Valley of lived in somewhat better made wattle-and-daub Jeffrey P. Blomster,1* Hector Neff,2 Michael D. Glascock3 huts than other villagers (6), San Lorenzo leaders probably lived in the so-called Red The first Mesoamerican civilization, the Gulf Coast Olmec, is associated with Palace, where they used basalt brought in hierarchical society, monumental art, and an internally consistent ideology, ex- from 60 km away to provide columns, step pressed in a distinct style and salient iconography. Whether the Olmec style coverings, and drainage channels for their arose in just one area or emerged from interactions among scattered contem- residence. Rather than engage in the typo- poraneous societies remains controversial. Using elemental analysis, we de- logical argument as to whether the Olmec termined the regional clay sources of 725 archaeological ceramic samples from constituted a state or a chiefdom, we simply across Mesoamerica. Exported Olmec-style ceramics originated from the San note the more complex level of sociopolitical Lorenzo region of the Gulf Coast, supporting Olmec priority in the creation organization achieved by the Olmec com- and spread of the first unified style and iconographic system in Mesoamerica. pared with that of other contemporaneous societies (7). More than 3000 years ago, the first civiliza- Olmec style. Two views have emerged on the The ‘‘San Lorenzo horizon’’ from 3150 tion in ancient Mexico, the Olmec, coalesced relationship of the Olmec with other contem- to 2800 yr B.P. (1200 to 850 B.C.E., or 1350 along the Mexican Gulf Coast. Much of mod- poraneous cultures, characterized as the to 1000 calibrated B.C.E.) refers to the ear- ern Mexico, along with Guatemala, Belize, ‘‘mother culture’’ and ‘‘sister culture’’ liest spread of Olmec-style ceramic vessels Honduras, and El Salvador, constitutes the cul- perspectives (4). Proponents of the mother- and figurines across Mesoamerica (3, 8, 9). tural region known as Mesoamerica (Fig. 1). culture perspective see the Olmec as more As defined at San Lorenzo (9), some Olmec Elements of the Olmec art style and associ- sociopolitically complex than coeval cultures, ceramic types (such as Calzadas Carved ated iconography—the first coherent icono- playing a central role in the dissemination of pottery) have excised designs that may ex- graphic system in Mesoamerica—that emerged key elements of Mesoamerican civilization. press ideological concepts through iconog- were depicted on stone monuments in the Gulf Sister-culture advocates characterize the Ol- raphy (Fig. 2, A and B). Two additional Coast, as well as on portable ceramic objects mec as at the same sociopolitical level as San Lorenzo horizon pottery types have a that have been found at sites throughout other contemporaneous chiefdoms; while fine kaolin clay body: Xochiltepec White Mesoamerica (1, 2). Because this iconography engaged in competitive interaction, the Ol- and Conejo Orange-on-White (Fig. 2, C and has been linked with the dissemination of the mec had no priority in the development of the D). Xochiltepec White has long been rec- social, political, and religious institutions of Olmec style and its distribution. Here, we ognized as a possible export ware across the Olmec, analyzing its origin and spread is endeavor to move beyond this dichotomy, Mesoamerica, although of uncertain prove- central to understanding the development of which obscures understanding the complex nance (6, 9). complex society in Mesoamerica. The Ol- nature of the Olmec and its relationships The lack of the full repertoire of Olmec- mec style_s appearance at select sites outside with its neighbors. style motifs at any one site and the apparent the Gulf Coast has been used to argue that it In terms of Early Formative complexity, regional differences in expression of the Olmec arose in one area only and was exported or, recent investigations by Ann Cyphers and style have been cited as evidence that each alternatively, emerged from interactions among scattered regional chiefdoms (3). Here, we use instrumental neutron activation analysis (INAA) on archaeologically exca- vated or collected ceramics to fingerprint their provenience and exchange during the later half of the Early Formative period EÈ3450 to 2850 years before the present (yr B.P.), or 1500 to 900 B.C.E.^. The Olmec and Olmec style in Meso- america. Not all features referred to as the Olmec style may be linked with the archae- ological Gulf Coast Olmec; thus, we distin- guish between the terms Olmec culture and

1Department of Anthropology, George Washington University, Washington, DC 20052, USA. 2Department of Anthropology, California State University, Long Beach, CA 90840, USA. 3Research Reactor Center, University of Missouri, Columbia, MO 65211, USA. *To whom correspondence should be addressed. Fig. 1. Map of Early Formative regional centers, showing modern national borders and Mexico City E-mail: [email protected] for reference. Dashed lines show approximate boundaries of Mesoamerica.

1068 18 FEBRUARY 2005 VOL 307 SCIENCE www.sciencemag.org R ESEARCH A RTICLES region contributed to the style, with no cen- with little prior occupation, grew to be as S2). Samples of raw clay include 160 clay tral origin of imagery (3). Ceramic data from large as contemporaneous San Jose´ Mogote samples from San Lorenzo, 25 from Oaxaca Oaxaca have shaped this debate for the past during the San Lorenzo horizon, serving as (12–14), 203 from the Mazata´n zone, 244 from 30 years, as archaeologists have used stylis- the center of a small chiefdom comparable to the Basin of Mexico, and 196 samples of raw tic analysis, as well as limited microscopic that in the Valley of Oaxaca (7). clay and modern/historic examples of pot- study of select sherds, to assert that villagers Ceramic data. Since the mid-1990s, a tery collected from pottery-making towns in created their own local set of Olmec-style wide variety of Early Formative ceramic the Valley of Oaxaca. Our data have been symbols (6, 10). This model concludes that, samples from regions of Mesoamerica have compared with an additional 527 archaeo- rather than importing Gulf Coast pottery and been submitted to the University of Missouri logical samples from later sites in the Valley iconography, Valley of Oaxaca potters pro- Research Reactor (MURR) for INAA (7, 12). of Oaxaca. duced and exported Olmec-style symbols made Although the 80 samples detailed in table S1 Clays are useful in sourcing analysis be- in distinctive gray-ware pottery, referred to as come primarily from Etlatongo and San cause they reflect the composition of local Delfina Fine Gray and Leandro Gray, across Lorenzo, the larger database includes a total geology. The clay available for pottery has Mesoamerica; recipients of this Oaxaca-made of 725 sourced archaeological pottery been documented in the three regions on pottery supposedly included the Gulf Coast samples from Tlapacoya (Basin of Mexico), which we focus: San Lorenzo, the Valley of Olmec (6, 11). To test this model, the first San Jose´ Mogote and other Valley of Oaxaca, and the Nochixtla´nValley(9, 13–16). author excavated Early Formative ceramics Oaxaca sites, Laguna Zope (Pacific Coast, Located in the Isthmian Saline Basin, San at the site of Etlatongo, in the Nochixtla´n Isthmus of ), Paso de la Amada Lorenzo sits on a salt dome located in Mio- Valley of the Mixteca Alta, northwest of the and other Mazata´n area sites (Soconusco re- cene sedimentary rocks, adjacent to Pleisto- Valley of Oaxaca (Fig. 1). Previously ex- gion, Pacific Coast of Chiapas), and San cene and Holocene alluvium deposited during plored by Roberto Za´rate Mora´n, the site, Isidro (Chiapas Central Depression) (table flooding. The sedimentary rocks provide ex- cellent pottery clays, including white-firing Fig. 2. Profile and exterior views clay (kaolin). Precambrian metamorphic rock, of Olmec pottery types pro- Cretaceous limestone, and Miocene volcanic duced at San Lorenzo. (A and B) ignimbrite are exposed in the Valley of Calzadas Carved pottery. [(A) is Oaxaca, along with Quaternary alluvium sample BLM060.] (C) Xochiltepec derived from these rocks. Clay was likely White (sample BLM042). (D) Conejo Orange-on-White (sam- obtained from weathered Precambrian ple BLM041). Adapted from (9); gneiss, schists, and the associated alluvium; drawings courtesy of M. Coe. kaolinite clays appear in scattered pockets (6). Similar rocks are exposed in the Nochixtla´n Valley, although Cretaceous lime- stone is more abundant. Tertiary volcanic rocks, alluvial clays, and residual clays (with a high calcium carbonate content formed from the decomposition of limestone parent materials) are available in the Etlatongo vicinity. All 80 pottery samples in table S1 have been recorded with photography and/or drawings, with all Etlatongo (n 0 42) and San Lorenzo (n 0 21) samples from well- documented archaeological contexts. Most of the ceramics in the larger database (table S2) Fig. 3. Olmec-style pottery exca- have not been illustrated, and some lack vated at Etlatongo. (A) Profile archaeological provenience information (13). and exterior of a bowl made with The sampling procedure was not random. San Lorenzo clay (sample Sherds were selected that would show both BLM003). (B)Reconstructionof a locally produced neckless jar local production and possible imports, (sample BLM006). including samples similar to Xochiltepec White and Conejo Orange-on-White. Thus, Olmec-style sherds are overrepresented at most sites in the sample. In selecting samples from Etlatongo, special attention was paid to the gray wares that may have been exported by the Valley of Oaxaca (Fig. 3A). Analysis. Sample preparation, INAA, and quantitative techniques (17, 18) are detailed in online materials (19). We used a variety of pattern-recognition techniques to discern compo- sitionally homogeneous groups in the data and then tested the effectiveness of elemental dis- crimination between the groups by calculating individual ceramics and clay-sample Mahala-

www.sciencemag.org SCIENCE VOL 307 18 FEBRUARY 2005 1069 R ESEARCH A RTICLES nobis distances from all group centroids. All Fig. 4. Bivariate plot of Mahalanobis distances were cross-validated; chromium and tanta- that is, each specimen was removed from lum concentrations in its presumed group before calculating its pottery samples as- signed to five well- own Mahalanobis distance from the group’s defined reference groups centroid (12, 20). This is a conservative ap- in the Early Formative proach to group evaluation that may exclude pottery database. El- true group members but minimizes the chances lipses represent 90% of erroneous attribution. Such a conservative confidence level for approach minimizes chances that Olmec-style group membership, with unique symbols for the sherds found outside the Gulf Coast will be samples of each group. incorrectly identified as imports rather than local production. Sherds that could fall into more than one group from the same region were not used in group definition (such as ‘‘San Lorenzo Local/White’’ in table S1); ex- cluding them from specific group membership means that they play no role in defining the centroid and variance-covariance structure of those groups. Fig. 5. Bivariate plot Production and consumption of Olmec- of tantalum and ytter- style ceramics. The majority (n 0 725) of bium shows the high more than 1000 Early Formative archaeo- concentration of tan- logical and clay samples separate into 15 talum in San Lorenzo regionally specific compositional groups; samples. Shaded/bold symbols represent the the remainder fall into 3 unassigned groups samples detailed in (table S3). These 15 groups are discrimi- table S1. Ellipses repre- nated well by the 32 elements retained for anal- sent 90% confidence ysis according to the Mahalanobis distance– level for group mem- based probabilities of group membership. bership, with unique Several elements, such as chromium, thorium, symbols for the sam- ples of each group. tantalum, cesium, and calcium, are particu- larly useful as discriminators of ceramics made in different regions. For example, the Mazata´n samples are all low in chromium, whereas the Tlapacoya samples are all enriched in chromium (Fig. 4). Three groups—Etlatongo-1, Etlatongo-2, and Oaxaca-1—can be associated with highland Oaxaca production. Although we have numer- virtually all Xochiltepec White and La Mina widespread distribution. Although some re- ous clay samples from the Valley of Oaxaca, White ceramics in the sample, were evidently gions produced local variants, these were not local production at Etlatongo was primarily produced with San Lorenzo clay. exported between regions; no non–Gulf Coast– sampled by selecting highly friable utilitarian One white-ware group, White-2, cannot produced white ware was exported to San wares that would be in need of frequent replace- be associated with a specific region. All nine Lorenzo. Xochiltepec White, which first ap- ment (7). The Etlatongo-1 and Etlatongo-2 White-2 samples in table S3 appear similar pears at San Lorenzo prior to the San Lorenzo groups represent the varied nature of the clays to Xochiltepec White pottery or a similar horizon (9), and Conejo Orange-on-White available in the Etlatongo vicinity. The data type. The chemistry of Xochiltepec White were Olmec products exported to select sites demonstrate that the same clays used for util- pottery clearly indicates that it was produced throughout Mesoamerica. itarian vessels could also be used for serving in the Gulf Coast, but all White-2 group In terms of decorated pottery, the San vessels with Olmec-style motifs (Fig. 3B). archaeological samples have been found in Lorenzo Local group contains most of the Potters thus produced Olmec-style ceramics the Valley of Oaxaca. Although this occur- Olmec-style ceramics with iconography in using local clays in both the Nochixtla´n and rence could represent a kaolin source in the the database, especially as expressed in gray- Oaxaca Valleys. Valley of Oaxaca, approximately half of the ware pottery. A problem did emerge early in Three groups—San Lorenzo White, San Lo- specimens assigned to the group exceed 1% the analysis in segregating pottery produced renzo Local, and San Lorenzo Local/White— probability of membership in the San at San Lorenzo from certain Oaxaca gray clearly derive from San Lorenzo region clays, Lorenzo White group. Thus, there remains wares. Generally, Gulf Coast clays contain even though the finished vessels are found at the possibility that the clay may be derived low concentrations of calcium, whereas those villages across Mesoamerica. The ceramics from slightly anomalous clays local to the San from highland Oaxaca generally have high in the combined San Lorenzo Local/White Lorenzo region. concentrations. Some samples of Gulf Coast group were produced with San Lorenzo re- The results of the INAA demonstrate that pottery and clay, however, have calcium gion clays, but it was not possible to distin- the Olmec produced luxury white wares, concentrations similar to examples of later guish among these groups; a similar problem particularly Conejo Orange-on-White and Oaxacan gray wares, as well as ethnographic precludes sample BLM009 from being firmly Xochiltepec White, at San Lorenzo. Other samples of Valley of Oaxaca pots and clay placed in the San Lorenzo Local group. All regions included in the study evidently (12). A multivariate perspective resolves this examples of Conejo Orange-on-White, and imported this pottery, which explains its problem; when all 32 elements retained from

1070 18 FEBRUARY 2005 VOL 307 SCIENCE www.sciencemag.org R ESEARCH A RTICLES Table 1. Secure regional assignments for Early Formative pottery

Region as identified by INAA Archaeological context Gulf Valley of Nochixtla´n Valley of Chiapas Central Isthmus of Mazata´n Total Coast Mexico Depression Tehuantepec San Lorenzo (Gulf Coast) 203 0 0 0 0 0 0 203 Mazata´n (various sites) 23 177 0 0 0 0 0 200 Valley of Oaxaca (various sites) 12 0 42 0 0 0 0 54 Etlatongo (Nochixtla´n Valley) 35 0 0 26 0 0 0 61 Tlapacoya (Valley of Mexico) 17 0 0 0 87 0 0 104 San Isidro (Chiapas Central Depression) 1000041042 Laguna Zope (Isthmus of Tehuantepec) 30000 0 5861 725 the analysis are considered, the discrimination Xochiltepec White served as the inspiration sumers and emulators rather than exporters into groups is unambiguous. For example, just for local ceramic emulations throughout and innovators of Olmec-style motifs. a bivariate plot of tantalum and ytterbium Mesoamerica. None of the locally produced We propose that the San Lorenzo Olmec segregates archaeological samples from Olmec-style pottery in the sample from the played a central role in synthesizing a distinct Oaxaca, Etlatongo, and San Lorenzo (Fig. 5). Valley of Oaxaca, the Nochixtla´n Valley, the style and associated iconography, disseminating In Oaxaca, Delfina and Leandro Gray pots Valley of Mexico, the Isthmus of Tehuante- it across Mesoamerica. The mechanisms for with Olmec-style motifs have been interpreted pec, the Chiapas Central Depression, and the exchange of Olmec-style pottery and symbols as expressing local variants of the Olmec style Mazata´n region was exported to the Gulf probably varied on a regional basis. Rather manufactured in the Valley of Oaxaca and Coast (Table 1). Contra to the sister-culture than being imposed by the Olmec, these exported to other Formative centers (6, 11). perspective, our analysis implies that the symbols were received by people, probably Our data document the presence of compo- Olmec at San Lorenzo did not receive leaders, at regional centers, who used, sitionally different gray wares in the Valley foreign-made variants of the Olmec style; the manipulated, and reproduced them in differ- of Oaxaca; it seems that both local manu- Olmec were its sole exporters. Despite the ent ways. Exchange of these symbols formed facture and Olmec imports are represented size of the San Lorenzo sample (n 0 203), an important component of communication at sites such as San Jose´ Mogote. Valley of not a single analyzed sherd was imported. and negotiation between communities on both Oaxaca–produced gray wares were not ex- Rather than expressing regional styles from intra- and interregional levels; non–Gulf Coast ported to the nearby Nochixtla´n Valley; in- across Mesoamerica at San Lorenzo, the INAA groups may have moved Olmec-produced stead, Etlatongo imported Olmec-style gray data emphasize the diversity of Calzadas Carved objects between regions rather than locally wares from the more distant Gulf Coast pottery produced by the Olmec. Further- produced pottery. The local impact of this (Fig. 3B). Indeed, all nonlocally produced more, the conservative approach used to interaction on communities varied. While this Olmec-style gray pottery samples found out- assign members to the compositional groups interaction may have had a transformative ef- side the Gulf Coast appear to be San Lorenzo may underestimate the numbers of San fect on the local population of the Mazata´n exports (Table 1). Lorenzo–derived specimens exported out- region (22), the impact in the Nochixtla´nVal- The INAA data thus contradict the model side the Gulf Coast. Because sampling ley appears to be substantially different and that regional exports can be identified simply was not random, the data in Table 1 do less profound. Despite the high frequency of by style in assemblages of Olmec-style pot- not represent frequencies of imported pottery imported Olmec pottery in Table 1, the total tery (6). At least one supposed ‘‘Oaxaca-style’’ for each region. quantity of Olmec-style pottery, both imported Calzadas Carved gray ware excavated at San Our data present several implications for and locally made, probably never exceeded 5 Lorenzo (sample BLM050) was analyzed by understanding the Olmec and their relation- to 10% of the total ceramic assemblage ex- INAA; it was produced with San Lorenzo clay. ships with contemporaneous groups. Olmec cavated at Etlatongo (7). The Olmec produced and exported a wide pottery and symbols associated with the Olmec Olmec priority in the creation of the Olmec variety of Olmec-style iconography, often were valued by elites at some of the largest style emphasizes their important role in Early on gray-ware pottery, across Mesoamerica. chiefly centers in Early Formative Meso- Formative social developments throughout much Local production of Olmec-style motifs also america. It does not appear that these other of Mesoamerica. Understanding the Olmec role occurred in all regions involved in this in- regions contributed substantially to the as- in disseminating this style and associated ico- teraction. At Tlapacoya, a Basin of Mexico semblage of Olmec-style symbols employed nography invokes more dynamic models of inter- village known for pottery with elaborate beyond each region. Not only did the Olmec regional interaction that explore the interests profile views of cleft-headed creatures (21), not import any foreign pottery with Olmec- of both producers and consumers of Olmec- Olmec-style designs occur on gray pottery style symbols, none of the regions in Table 1 style pottery. This multifaceted interaction with both San Lorenzo (BLM039) and local exchanged pottery with each other. Etlatongo anticipates the increasingly complex nature of (BLM062) origins. Once again, none of the and San Jose´ Mogote, in adjacent regions of Olmec networks in the following Middle Form- decorated and undecorated gray pottery analyzed highland Oaxaca, received pottery from the ative Period, when La Venta became the Gulf from Tlapacoya was imported from Oaxaca. more distant San Lorenzo rather than from Coast’s largest Olmec center. Olmec and Early Formative interac- each other. This appears to be a stark repu- tion. The San Lorenzo Olmec produced diation of the view of mutual exchange and References and Notes both fine white-paste pottery and decorated production of these symbols, and it contra- 1. M. Coe, in Handbook of Middle American Indians, Olmec-style ceramics. The Olmec exported this dicts the claim that Oaxacan Delfina Fine gray Vol. 3: Archaeology of Southern Mesoamerica, part 2, pottery to other regional centers across Meso- pottery was as popular an export ware as R. Wauchope, G. Willey, Eds. (Univ. Texas Press, Austin, 1965), pp. 739–775. america, where many centers created local Xochiltepec White (6). The regions outside 2. E. Benson, B. de la Fuente, Eds., Olmec Art of Ancient variants of it; indeed, we suspect Gulf Coast the Gulf Coast appear to be primarily con- Mexico (National Gallery of Art, Washington, DC, 1996).

www.sciencemag.org SCIENCE VOL 307 18 FEBRUARY 2005 1071 R ESEARCH A RTICLES

3. R. Sharer, D. Grove, Eds., Regional Perspectives on the 14. M. Thieme, in Procesos de Cambio y Conceptualizacio´n 23. We thank the technicians at MURR and the ar- Olmec (Cambridge Univ. Press, New York, 1989). del Tiempo: Memoria de la Primera Mesa Redonda de chaeologists in Mexico who provided samples for this 4. D. Grove, J. World Prehistory 11, 51 (1997). Monte Alba´n, N. Robles Garcı´a, Ed. (Instituto Nacional study, as well as funding agencies that made their 5. A. Cyphers, in Olmec to Aztec: Settlement Patterns in de Antropologı´a e Historia, Mexico City, 2001), pp. collections possible. The support and permissions of the Ancient Gulf Lowlands,B.Stark,P.ArnoldIII,Eds. 341–349. the Instituto Nacional de Antropologı´a e Historia— (Univ. of Arizona Press, Tucson, AZ, 1997), pp. 96–114. 15. J. Lorenzo, Rev. Mex. Estud. Antropol. 16, 49 (1960). especially L. Mirambell, M. Serra Puche, J. Garcı´a- 6.K.Flannery,J.Marcus,Early Formative Pottery of the Valley 16. M. Kirkby, The Physical Environment of the Nochixtl´an Ba´rcena, E. Lo´pez Calzada, and N. Robles Garcia—have of Oaxaca, Mexico, Memoirs of the Museum of Anthro- Valley, Oaxaca, Vanderbilt University Publications in been invaluable. Funding for the 80 samples in pology 27 (Univ. of Michigan, Ann Arbor, MI, 1994). Anthropology 2 (Vanderbilt Univ., Nashville, TN, 1972). table S1 came primarily from the Charles J. MacCurdy 7. J. Blomster, Etlatongo: Social Complexity, Interac- 17. M. Glascock, in Chemical Characterization of Ceramic Endowment, Yale University; the non-Etlatongo sam- tion, and Village Life in the Mixteca Alta of Oaxaca, Pastes in Archaeology, H. Neff, Ed. (Prehistory Press, ples were provided by M. Winter and M. Coe. The Mexico (Wadsworth, Belmont, CA, 2004). Madison, WI, 1992), pp. 11–26. research at the MURR Archaeometry Lab has been 8. J. Blomster, Ancient Mesoamerica 9, 171 (1998). 18. H. Neff, in Modern Analytical Methods in Art and supported by NSF (no. SBR-9802366). We are grate- 9. M. Coe, R. Diehl, In the Land of the Olmec, Vol. 1: The Archaeology, E. Ciliberto, G. Spoto, Eds. (Wiley, New ful for the support of the Sainsbury Research Unit, Archaeology of San Lorenzo Tenochtitla´n (Univ. of York, 2000), pp. 81–134. University of East Anglia. We thank R. Diehl, A. Joyce, Texas Press, Austin, TX, 1980). 19. Materials and methods are available as supporting G. Lau, and an anonymous reviewer for comments. 10. K. Flannery, in Dumbarton Oaks Conference on the material on Science Online. Olmec, E. Benson, Ed. (Dumbarton Oaks, Washing- 20.R.Bishop,H.Neff,inArchaeological Chemistry IV, Supporting Online Material ton, DC, 1968), pp. 79–117. R. Allen, Ed. (American Chemical Society, Washington, www.sciencemag.org/cgi/content/full/307/5712/1068/ 11. K. Flannery, J. Marcus, J. Anthropol. Archaeol. 19,1 DC, 1989), pp. 576–586. DC1 (2000). 21. P. Joralemon, A Study of Olmec Iconography, Studies Materials and Methods 12. H. Neff, M. Glascock, ‘‘Report on Instrumental Neutron in Pre-Columbian Art and Archaeology 7 (Dumbarton Tables S1 to S3 Activation Analysis of Olmec Pottery’’ (Research Re- Oaks, Washington, DC, 1971). References actor Center, Univ. of Missouri, Columbia, MO, 2002). 22. J. Clark, M. Pye, in Olmec Art and Archaeology in 13. R. Herrera, H. Neff, M. Glascock, J. Elam, J. Archaeol. Mesoamerica, J. Clark, M. Pye, Eds. (National Gallery 16 November 2004; accepted 6 January 2005 Science 26, 967 (1999). of Art, Washington, DC, 2000), pp. 217–251. 10.1126/science.1107599

some cases, these issues have been studied Whole-Genome Patterns of in limited genomic intervals, but comprehen- sive genomic analyses have not been possible. Common DNA Variation in Genome-wide association studies to iden- tify alleles contributing to complex traits of medical interest are currently performed with Three Human Populations subsets of common SNPs, and thus they rely David A. Hinds,1 Laura L. Stuve,1 Geoffrey B. Nilsen,1 on the expectation that a disease allele is Eran Halperin,2 Eleazar Eskin,3 Dennis G. Ballinger,1 likely to be correlated with an allele of an assayed SNP. Although studies have shown Kelly A. Frazer,1 David R. Cox1* that variants in close physical proximity are often strongly correlated, this correlation Individual differences in DNA sequence are the genetic basis of human structure, or linkage disequilibrium (LD), is variability. We have characterized whole-genome patterns of common human complex and varies from one region of the DNA variation by genotyping 1,586,383 single-nucleotide polymorphisms genome to another, as well as between (SNPs) in 71 Americans of European, African, and Asian ancestry. Our results different populations (5, 6). Selection of a indicate that these SNPs capture most common genetic variation as a result maximally informative subset of common of linkage disequilibrium, the correlation among common SNP alleles. We SNPs for use in association studies is neces- observe a strong correlation between extended regions of linkage dis- sary to provide sufficient power to assess the equilibrium and functional genomic elements. Our data provide a tool for causal role of common DNA variation in exploring many questions that remain regarding the causal role of common complex human traits. Although a large frac- human DNA variation in complex human traits and for investigating the tion of all common human SNPs are avail- nature of genetic variation within and between human populations. able in public databases, lack of information concerning SNP allele frequencies and the Single-nucleotide polymorphisms (SNPs) are single-base variants, most of which exist in correlation structure of SNPs within and the most abundant form of DNA variation in only a single individual. between human populations has made it the human genome. It has been estimated The relationship between DNA variation difficult to select an optimal subset. that there are È7 million common SNPs with and human phenotypic differences (such as Here we examine the SNP allele frequen- a minor allele frequency (MAF) of at least height, eye color, and disease susceptibility) cies and patterns of LD between 1,586,383 5% across the entire human population (1). is poorly understood. Although there is evi- SNPs distributed uniformly across the human Most common SNPs are to be found in most dence that both common SNPs and rare genome in unrelated individuals of Europe- major populations, although the frequency of variants contribute to the observed variation an, African, and Asian ancestry. Our primary any allele may vary considerably between in complex human traits (3, 4), the relative aim was to create a resource for further populations (2). An additional 4 million contribution of common versus rare variants investigation of the structure of human SNPs exist with a MAF between 1 and 5%. remains to be determined. The structure of genetic variation and its relationship to In addition, there are innumerable very rare genetic variation between populations and its phenotypic differences. relationship to phenotypic variation is un- A dense SNP map. To characterize a 1 Perlegen Sciences Inc., 2021 Stierlin Court, Mountain clear. Similarly, the relative contribution to panel of markers that would be informative View, CA 94043, USA. 2International Computer 3 complex human traits of DNA variants that in whole-genome association studies, we se- Science Institute, Berkeley, CA 94704, USA. Depart- ment of Computer Science and Engineering, Univer- alter protein structure by amino acid replace- lected a total of 2,384,494 SNPs likely to be sity of California–San Diego, La Jolla, CA 92093, USA. ment, versus variants that alter the spatial or common in individuals of diverse ancestry *To whom correspondence should be addressed. temporal pattern of gene expression without (7). We identified the majority (69%) of the E-mail: [email protected] altering protein structure, is unknown. In SNPs by performing array-based resequencing

1072 18 FEBRUARY 2005 VOL 307 SCIENCE www.sciencemag.org