Article

The timing and effect of the earliest human

arrivals in North America

https://doi.org/10.1038/s41586-020-2491-6 Lorena Becerra-Valdivia1,2 ✉ & Thomas Higham1

Received: 17 October 2019

Accepted: 26 May 2020 The peopling of the Americas marks a major expansion of humans across the planet.
Published online: xx xx xxxx However, questions regarding the timing and mechanisms of this dispersal remain,
and the previously accepted model (termed ‘Clovis-first’)—suggesting that the first
Check for updates
inhabitants of the Americas were linked with the Clovis tradition, a complex marked
by distinctive fluted lithic points1—has been effectively refuted. Here we analyse
chronometric data from 42 North American and Beringian archaeological sites using
a Bayesian age modelling approach, and use the resulting chronological framework to
elucidate spatiotemporal patterns of human dispersal. We then integrate these
patterns with the available genetic and climatic evidence. The data obtained show
that humans were probably present before, during and immediately after the Last
Glacial Maximum (about 26.5–19 thousand years ago)2,3 but that more widespread
occupation began during a period of abrupt warming, Greenland Interstadial 1 (about
14.7–12.9 thousand years before ad 2000)4. We also identify the near-synchronous
commencement of Beringian, Clovis and Western Stemmed cultural traditions, and
an overlap of each with the last dates for the appearance of 18 now-extinct faunal
genera. Our analysis suggests that the widespread expansion of humans through
North America was a key factor in the extinction of large terrestrial mammals.

Until recently, the prevailing paradigm1 for the initial dispersal of (Supplementary Information, Supplementary Table 1). This approach
humans into the Americas held that the first American individuals enables the incorporation of absolute chronometric data (in this case,
were big-game hunters who entered the continent about 13 thousand radiocarbon and luminescence ages) with stratigraphic and other rela-
years ago (ka) from Asia via Beringia, moving southwards through an tive age information (Methods). The sites we have analysed include
ice-free corridor between the Laurentide and Cordilleran Ice Sheets. cultural components that fall, according to the literature, within one
This paradigm proposed that, once below the 48th parallel north, of three technological traditions (Clovis, Western Stemmed or Bering-
these human groups developed a material tradition (named Clovis) ian) (Supplementary Information) or—as defined here—two exclusively
that spread across North America, and which dates to between 13,250 temporal categories (pre-Clovis and ‘Clovis-coeval’) (Fig. 1). We gener-
and 12,800 calibrated years before present (ad 1950) (cal. bp)5. This ated probability density functions that correspond to temporal start
narrative, known as the ‘Clovis-first’ model, was widely accepted for boundaries or age estimates for each site and tradition from the models
most of the twentieth century as it effectively answered most of the (Methods, Supplementary Information, Supplementary Table 2), which
questions associated with the peopling of the Americas—when, why are reported below at a 95.4% confidence or credible interval. We also
and from where1. However, the Clovis-first model has been refuted used kernel density estimate methods, which estimate the underlying
by new archaeological and chronometric data6–8 that demonstrate distribution of a dataset, to visualize the spatial and temporal distribu-
the existence of sites that predate Clovis (often termed ‘pre-Clovis’ tion of the data (Methods). We used the Greenland ice-core timescale
or ‘older-than-Clovis’6,9) as well as of distinct technological industries (GICC05)20 to place our results in the context of past climatic changes,
that occur coevally10–14. An earlier dispersal route along the Pacific particularly the period that comprises Greenland Interstadial 1 (GI-1)
Coast is currently considered the most likely means by which people and Greenland Stadial 1 (GS-1)—beginning at about 14.7 thousand years
moved into the Americas15,16, although this not universally accepted17. before ad 2000 and ending at about 11.7 thousand years before ad
To understand the pattern of initial human dispersals in a more 20004 (Figs. 2, 3). Although GICC05 and the current radiocarbon dat-
refined manner, we have used Bayesian statistical methods to analyse ing calibration curve (IntCal1321) are not synchronous, between 14,000
chronometric data from a large number of North American and Ber- and 11,000 cal. bp the estimated offset is less than around 50 years22.
ingian archaeological sites. Our results enable us to propose a new Given the temporal resolution of this study, the effect of this offset is
chronology-based model for the peopling of North America. therefore negligible.
We built high-precision chronometric models using Bayesian age Modelled start boundaries for pre-Clovis sites show the earli-
modelling on the OxCal 4.3 platform18,19 on the basis of archaeological est evidence for cultural occupation in stratigraphic component C
and chronometric information obtained from 42 archaeological sites of Chiquihuite Cave (Mexico) at 33,150–31,405 cal. bp, before the

Oxford Radiocarbon Accelerator Unit, Research Laboratory for Archaeology and History of Art, University of Oxford, Oxford, UK. 2Chronos 14C-Cycle Facility, SSEAU, University of New
1

South Wales, Sydney, New South Wales, Australia. ✉e-mail: l.becerravaldivia@unsw.edu.au

Nature | www.nature.com | 1
Article

1
Beringia
3
4
9, 10
5–8 11

Co e Sh
Ic
rdi ee
Laurentide

lle t
2

ran
Ice Sheet

12 14
13 16
15 17 22, 23
18 21
20 28
19 24 25 26 27
Pacific Ocean 30 29
33 31
35 36, 37 32
34 38, 39 41
40 Atlantic Ocean
42

1. Berelekh 11. Swan Point 21. Lange/Ferguson 31. Domebo

2. Ushki 12. Meay Channel I 22. Hebior 32. Aubrey
3. Mesa 13. Manis 23. Schaefer 33. Blackwater Draw
4. Bluefish Caves 14. Wally’s Beach 24. Bonneville Estates 34. Lubbock Lake
5. Dry Creek 15. Cooper’s Ferry 25. Dent 35. Arlington Springs
6. Walker Road 16. Lindsay 26. Sheriden Cave 36. Murray Springs
7. Moose Creek 17. Anzick 27. Meadowcroft Rockshelter 37. Lehner
8. Owl Ridge 18. Colby 28. Shawnee–Minisink 38. Gault
9. Broken Mammoth 19. Paisley Caves 29. Cactus Hill 39. Debra L. Friedkin
10. Mead 20. Buhl 30. Jake Bluff 40. El Fin del Mundo
Beringian tradition Clovis 41. Page–Ladson
Clovis-coeval WST 42. Chiquihuite Cave
Pre-Clovis Containing pre-Clovis and Clovis/WST components Beringian coastline during the LGM

Fig. 1 | Map showing the location of the 42 archaeological sites included in this study. The extent of the ice-free corridor at 12,500 cal. bp is according to ref. 42,
and the Beringian coastline during the LGM is approximated after ref. 77. WST, Western Stemmed tradition. Scale bar, 4,000 km.

Last Glacial Maximum (LGM) (from 26.5 to 20–19 ka2,3). Several sites Page–Ladson (14,710–14,450 cal. bp), Lindsay (14,625–13,945 cal. bp)
appear to be occupied later, during or immediately after the LGM. and Manis (13,795–13,745 cal. bp) fall within or near GI-1. Chronometric
These include Gault (26,435–17,385 cal. bp), Meadowcroft Rockshelter data (n = 171 dates) from these 14 pre-Clovis components yielded a
(24,335–18,620 cal. bp) and Cactus Hill (20,585–18,970 cal. bp) (Fig. 2a). distribution with a peak centred on around 14,250 cal. bp, which rep-
In eastern Beringia, Bluefish Caves is represented by a single date resents the bulk of the chronometric evidence (Fig. 2b). For the later
obtained on a humanly modified bone sample (24,035–23,310 cal. bp) Beringian, Western Stemmed and Clovis lithic traditions, the modelled
dating squarely to the LGM. Start boundaries for stratigraphic com- start boundaries we obtained suggest that they began at 14,955–13,895,
ponent B of Chiquihuite Cave (16,605–15,615 cal. bp), Cooper’s Ferry 14,860–13,065 and 14,210–13,495 cal. bp, respectively (Fig. 3, Extended
(16,560–15,285 cal. bp) and Debra L. Friedkin (16,315–14,660 cal. bp) Data Figs. 1–3).
suggest occupations at these sites began after the LGM and near to Apart from pre-LGM evidence (currently limited to stratigraphic
GI-1—a warmer period of sudden, short-lived climatic fluctuations4. component C of Chiquihuite Cave), our analysis shows that several
Following these sites, age estimates for Hebior (15,615–13,975 cal. bp), archaeological sequences in North America—Bluefish Caves, Gault,
Schaefer (15,020–13,710 cal. bp), Paisley Caves (14,755–13,780 cal. bp), Meadowcroft Rockshelter and Cactus Hill—begin during or immediately

a b
–35
δ18O

–40

–45
Manis (date function)
Lindsay (date function)
Page–Ladson (start boundary)
Probability density

Paisley Caves (date function) 0.04

Schaefer (date function)
Hebior (date function)
Debra L. Friedkin (start boundary) 0.02
Cooper’s Ferry (start boundary)
Chiquihuite Cave SC-B (start boundary)
Cactus Hill (start boundary)
Bluefish Caves (calibrated date) 0
Meadowcroft Rockshelter (start boundary)
Gault (start boundary)
Chiquihuite Cave SC-C (start boundary)
90,000 80,000 70,000 60,000 50,000 40,000 30,000 20,000 10,000 35,000 30,000 25,000 20,000 15,000 10,000 5,000
Date (cal. BP) Date (cal. BP)

Fig. 2 | Start boundaries or age estimates for pre-Clovis sites or of chronometric data (n = 171 dates). The light grey crosses show the median
components, and a summarized distribution of chronometric data. a, Start values of the calibrated date ranges, and the black crosses show the medians of
boundaries or age estimates for 14 pre-Clovis sites or components. The LGM the marginal posterior distributions for each event from the kernel density
span is according to ref. 78, and is noted with a yellow band; GI-1 (dark grey band) estimate model analysis. Within this distribution, data are concentrated
and GS-1 (light grey band) follow ref. 4. δ18O data are according to the Greenland around a single peak at about 14,250 cal. bp. These results suggest that humans
ice-core timescale (GICC05)20. Brackets beneath each age estimate show 95.4% were initially present in North America before, during or immediately after the
confidence interval. SC, stratigraphic component. b, Summarized distribution LGM.

2 | Nature | www.nature.com
 –35
δ18O

–40
0.006

Probability density
0.004
–45
0.002
LADs for 24 extinct mammal genera
0
Start of WST
Start of Clovis
Start of Beringian tradition
adition

45,000 40,000 35,000 30,000 25,000 20,000 15,000 10,000 50,000 40,000 30,000 20,000 10,000 0
Date (cal. BP) Date (cal. BP)

Fig. 3 | Start boundaries for the Clovis, Western Stemmed and Beringian crosses show the medians of the marginal posterior distributions for each event
traditions, and a summarized distribution of the dates for the last from the kernel density estimate model analysis. Data for the last appearance
appearance of 24 extinct mammal genera in North America. Left, start dates (LADs) of extinct fauna (n = 24) are entered as compiled in ref. 73.
boundaries for the Clovis, Western Stemmed and Beringian traditions. The Our results suggest that the onset of the three traditions is largely coincident
LGM span is according to ref. 78, and is noted with a yellow band; GI-1 (dark grey with GI-1, a globally warmer period of marked, short-lived climate changes4.
band) and GS-1 (light grey band) follow ref. 4. δ18O data are according to the Moreover, the highest density peak for the kernel density estimate model
Greenland ice-core timescale (GICC05)20. Brackets beneath each age estimate overlaps with the start of the Clovis and Western Stemmed traditions, and
show 95.4% confidence interval. In the summarized distribution (right), the coincides with the interface between GI-1 and GS-1. This suggests that human
light grey crosses show the median values of the calibrated dates, and the black expansion was strongly involved in Pleistocene extinctions in North America.

after the LGM and, south of the 48th parallel north, concentrate in the indistinguishable (Supplementary Information section 5) and occur
east (Fig. 1, Extended Data Fig. 4a, b). Although the global climate was near-synchronously, largely coinciding with GI-1 (Fig. 3) and a peak
generally cooler and drier during the LGM, areas to the south of the in the modelled chronometric data (Extended Data Fig. 4e, f). This
North American ice sheets were relatively open and temperate23–25, suggests a probable increase in population density at this time, which
with evidence for high hydrological and ecological variability26–31. The is supported by mitochondrial49, Y chromosome50 and autosomal51
distribution of early mid-latitude sites has led some to argue that North evidence for marked population growth at about 16–15 ka. In addi-
America was originally colonized by southwestern European individu- tion, the age estimates probably follow or coincide with a split of the
als who arrived via a transatlantic ice corridor during the LGM; the ancestral Native American lineage into two branches45,52–55 (Southern
so-called Solutrean hypothesis32. The connection to Europe lies in and Northern), which probably occurred outside of Eastern Beringia
purported typological and manufacturing similarities between Solu- and south of the North American ice sheets45,46,52 around 17.5–14.6 ka46.
trean (appearing at about 25,000–24,000 cal. bp33) and Clovis lithic Technologically, the relationship between pre-Clovis and Clovis
technologies, as well as evidence for pre-Columbian genetic admix- traditions is uncertain9,47, whereas the Western Stemmed tradition has
ture between ancestral Native American and West Eurasian popula- previously been linked with several pre-Clovis7,8,15 and circum-Pacific
tions. However, this hypothesis has been rejected on technological sites56,57 owing to perceived similarities in stemmed point technolo-
and genetic grounds34,35. If transatlantic migration is set aside and an gies. However, recent evidence from the Debra L. Friedkin site has
Asian origin assumed, the antiquity and distribution of the early sites been argued to show the progression from stemmed to fluted (Clovis)
suggest that the initial crossing of the 48th parallel north occurred technology, with a triangular lanceolate point noted as a possible link
either (i) during the later part of Marine Isotope Stage 3 (57–29 ka)36, between the two8. Genetically, the connection between pre-Clovis
when ice and sea level estimates37–39 indicate that land passage through populations and the makers of Clovis and Western Stemmed traditions
Beringia was unlikely or interrupted, and an ice-free corridor between requires further research. This is because genetic information from
the Laurentide and Cordilleran Ice Sheets was probably present39 (with other Clovis-coeval or pre-Clovis archaeological sites has yet to be
evidence of terrestrial landscapes occurring between 48 and 40 ka40) obtained, with the exception of the pre-Clovis component at Paisley
or (ii) during the LGM terminus, when the Bering land bridge was viable Caves (where mitochondrial DNA data correspond to Native American
but the ice-free corridor was inaccessible41,42. Both possibilities sug- founding haplogroups58) and Anzick, a burial site that represents Clovis
gest the earliest arrivals to North America had some degree of littoral and the Southern Native American lineage53. However, it is likely that
adaptation. The second scenario would have required movement along these groups and the initial arrivals are somehow related—otherwise,
the Pacific Coast, perhaps before the Cordilleran Ice Sheet reached its a minimum of two Pleistocene-aged migrations should be expected.
maximum extent (between about 20 and 17 ka43). It is also more compat- Given our results, if there were a later dispersal event into the Americas,
ible with current genetic findings, which suggest that ancestral Native this would most probably predate GI-1 and estimates for the earliest
American populations experienced genetic isolation in eastern Beringia bioavailability of the ice-free corridor around 13,000 cal. bp59,60, perhaps
during the LGM44 and diverged from Ancient Beringian populations at occurring after the western retreat of the Cordilleran Ice Sheet around
22–18.1 ka45,46. This assumes the pre-Clovis occupants of Gault, Mead- 18–17 ka43,61,62 and the Alaska Peninsula Glacier Complex at about 17 ka,
owcroft Rockshelter and Cactus Hill were of Native American ancestry. as well as the likely establishment of productive ecosystems in the North
Evidence from stratigraphic component C of Chiquihuite Cave suggests Pacific coast by about 17 ka43.
the presence of an earlier human group, but their genetic ancestry is Humans arriving in North America have previously been linked
as yet unknown47. Once in mid-latitude North America, the paucity with the extinction of 37 faunal genera, in what is known as the over-
(only 12 archaeological sites), distribution (Fig. 1) and characteristics kill hypothesis63. Depending on the analysis, more recent studies are
of pre-Clovis evidence suggest that humans had unique adaptational either in support64 or suggest other factors—such as climate, ecological
behaviours and were widely dispersed. Lithic industries at Chiquihuite change and an extraterrestrial impact—as main or joint contributors65–72.
Cave, Gault, Meadowcroft Rockshelter and Cactus Hill, for example, Our results show that human presence in the continent precedes the
are largely—although perhaps not entirely48—unrelated9,47. majority of the dates for the last appearances of extinct genera in
Bayesian age modelling shows that the start dates of the Ber- North America73, including Camelops, Cuvieronius, Equus, Mammut
ingian, Western Stemmed and Clovis traditions are statistically and Mammuthus. When these are summarized in a probability density

Nature | www.nature.com | 3
Article
distribution, the highest peak (containing 18 of the 24 dates included) 15. Davis, L. G. et al. Late Upper Paleolithic occupation at Cooper’s Ferry, Idaho, USA shows
Americas settled before ~16,000 years ago. Science 365, 891–897 (2019).
occurs during the boundary between GI-1 and GS-1, and overlaps with
16. Braje, T. J., Dillehay, T. D., Erlandson, J. M., Klein, R. G. & Rick, T. C. Finding the first
the start boundaries obtained for Clovis and Western Stemmed tradi- Americans. Science 358, 592–594 (2017).
tions (Fig. 3). This raises the distinct possibility that widespread human 17. Potter, B. A. et al. Current evidence allows multiple models for the peopling of the
Americas. Sci. Adv. 4, eaat5473 (2018).
expansion in population and space was a key factor in the extinction
18. Bronk Ramsey, C. Development of the radiocarbon program OxCal. Radiocarbon 43,
of large terrestrial mammals. To better understand the relationship 355–363 (2001).
between human arrival, faunal extinctions and the climate changes 19. Bronk Ramsey, C. Bayesian analysis of radiocarbon dates. Radiocarbon 51, 337–360
(2009).
occurring at the time, we require improved demographic histories for 20. Rasmussen, S. O. et al. A new Greenland ice core chronology for the last glacial
each extinct genus and species, as well as more robust chronometric termination. J. Geophys. Res. 111, D06102 (2006).
data to improve Bayesian models74. Results from the latter would allow 21. Reimer, P. J. et al. IntCal13 and Marine13 radiocarbon age calibration curves 0–50,000
years cal BP. Radiocarbon 55, 1869–1887 (2013).
formal integration with the archaeological age estimates in this study 22. Adolphi, F. et al. Connecting the Greenland ice-core and U/Th timescales via cosmogenic
as well as their combined, quantitative analysis. radionuclides: testing the synchroneity of Dansgaard–Oeschger events. Clim. Past 14,
Our study of the timing of initial human dispersals into Beringia 1755–1781 (2018).
23. Ray, N. & Adams, J. A GIS-based vegetation map of the world at the last glacial maximum
and North America suggests that people were present in different set- (25,000–15,000 BP). Internet Archaeol. 11, https://doi.org/10.11141/ia.11.2 (2001).
tings before, during and immediately following the LGM, before the 24. Jackson, S. T. et al. Vegetation and environment in Eastern North America during the Last
commencement of more-widespread occupation and human popu- Glacial Maximum. Quat. Sci. Rev. 19, 489–508 (2000).
25. Williams, J. W. Variations in tree cover in North America since the Last Glacial Maximum.
lation growth during GI-1. Pre-LGM evidence is currently limited to Global Planet. Change 35, 1–23 (2003).
one archaeological site (Chiquihuite Cave). If population continuity 26. Lyle, M. et al. Out of the tropics: the Pacific, Great Basin lakes, and late Pleistocene water
is assumed, this pattern is consistent with phases of human explora- cycle in the western United States. Science 337, 1629–1633 (2012).
27. Goebel, T., Hockett, B., Adams, K. D., Rhode, D. & Graf, K. Climate, environment, and
tion and colonization, and the degree of genetic structure already humans in North America’s Great Basin during the Younger Dryas, 12,900–11,600
present in North America by GI-1. However, the biocultural relationship calendar years ago. Quat. Int. 242, 479–501 (2011).
between the humans represented by pre-Clovis sites and later North 28. Menking, K. M., Anderson, R. Y., Shafike, N. G., Syed, K. H. & Allen, B. D. Wetter or colder
during the Last Glacial Maximum? Revisiting the pluvial lake question in southwestern
American and Beringian traditions is largely unknown. The analysis North America. Quat. Res. 62, 280–288 (2004).
of ancient human DNA from sediments75 has the potential to increase 29. Kirby, M. E. et al. A late Wisconsin (32–10k cal a BP) history of pluvials, droughts and
our knowledge and shed light on these questions. Finally, although this vegetation in the Pacific south-west United States (Lake Elsinore, CA). J. Quat. Sci. 33,
238–254 (2018).
study focused on Beringia and North America, continued investiga- 30. Ibarra, D. E. et al. Warm and cold wet states in the western United States during the
tions in Central and South America—for which Late Pleistocene data Pliocene–Pleistocene. Geology 46, 355–358 (2018).
are comparatively limited (as discussed in a special issue of Quaternary 31. Feakins, S. J., Wu, M. S., Ponton, C. & Tierney, J. E. Biomarkers reveal abrupt switches in
hydroclimate during the last glacial in southern California. Earth Planet. Sci. Lett. 515,
International76)—should allow for the chronological assessment of 164–172 (2019).
local archaeological sites6,10,14 and the development of continent-wide 32. Stanford, D. J. & Bradley, B. A. Across Atlantic Ice: The Origin of America’s Clovis Culture
spatiotemporal models. Future research will enable our chronological (Univ of California Press, 2013).
33. Aubry, T. & Almeida, M. Analyse critique des bases chronostratigraphiques de la
model to be tested, and its precision and accuracy improved as new structuration du Solutréen. Le Solutréen 40, 37e52 (2013).
data are integrated. 34. Eren, M. I., Patten, R. J., O’Brien, M. J. & Meltzer, D. J. Refuting the technological
cornerstone of the Ice-Age Atlantic crossing hypothesis. J. Archaeol. Sci. 40, 2934–2941
(2013).
35. Raff, J. A. & Bolnick, D. A. Does mitochondrial haplogroup X indicate ancient trans-Atlantic
Online content migration to the Americas? A critical re-evaluation. PaleoAmerica 1, 297–304 (2015).
36. Lisiecki, L. E. & Raymo, M. E. A Pliocene–Pleistocene stack of 57 globally distributed
Any methods, additional references, Nature Research reporting sum-
benthic δ18O records. Paleoceanography 20, PA1003 (2005).
maries, source data, extended data, supplementary information, 37. Spratt, R. M. & Lisiecki, L. E. A Late Pleistocene sea level stack. Clim. Past 12, 1079–1092
acknowledgements, peer review information; details of author con- (2016).
38. Pico, T., Mitrovica, J. X., Ferrier, K. L. & Braun, J. Global ice volume during MIS 3 inferred
tributions and competing interests; and statements of data and code
from a sea-level analysis of sedimentary core records in the Yellow River Delta. Quat. Sci.
availability are available at https://doi.org/10.1038/s41586-020-2491-6. Rev. 152, 72–79 (2016).
39. Batchelor, C. L. et al. The configuration of Northern Hemisphere ice sheets through the
Quaternary. Nat. Commun. 10, 3713 (2019).
1. Meltzer, D. J. The Great Paleolithic War: How Science Forged an Understanding of 40. Dalton, A. S., Finkelstein, S. A., Barnett, P. J. & Forman, S. L. Constraining the Late
America’s Ice Age Past (Univ. Chicago Press, 2015). Pleistocene history of the Laurentide Ice Sheet by dating the Missinaibi Formation,
2. Mix, A. C., Bard, E. & Schneider, R. Environmental processes of the Ice Age: land, oceans, Hudson Bay Lowlands, Canada. Quat. Sci. Rev. 146, 288–299 (2016).
glaciers (EPILOG). Quat. Sci. Rev. 20, 627–657 (2001). 41. Pico, T., Mitrovica, J. X. & Mix, A. C. Sea level fingerprinting of the Bering Strait flooding
3. Clark, P. U. et al. The Last Glacial Maximum. Science 325, 710–714 (2009). history detects the source of the Younger Dryas climate event. Sci. Adv. 6, eaay2935
4. Rasmussen, S. O. et al. A stratigraphic framework for abrupt climatic changes during the (2020).
Last Glacial period based on three synchronized Greenland ice-core records: refining and 42. Dyke, A. S. An outline of North American deglaciation with emphasis on central and
extending the INTIMATE event stratigraphy. Quat. Sci. Rev. 106, 14–28 (2014). northern Canada. Develop. Quat. Sci. 2, 373–424 (2004).
5. Waters, M. R. & Stafford, T. W. Jr. Redefining the age of Clovis: implications for the 43. Lesnek, A. J., Briner, J. P., Lindqvist, C., Baichtal, J. F. & Heaton, T. H. Deglaciation of the
peopling of the Americas. Science 315, 1122–1126 (2007). Pacific coastal corridor directly preceded the human colonization of the Americas. Sci.
6. Dillehay, T. D. Monte Verde, a Late Pleistocene Settlement in Chile: The Archaeological Adv. 4, eaar5040 (2018).
Context and Interpretation (Smithsonian Institution Press, 1997). 44. Tamm, E. et al. Beringian standstill and spread of Native American founders. PLoS ONE 2,
7. Williams, T. J. et al. Evidence of an early projectile point technology in North America at e829 (2007).
the Gault site, Texas, USA. Sci. Adv. 4, eaar5954 (2018). 45. Moreno-Mayar, J. V. et al. Early human dispersals within the Americas. Science 362,
8. Waters, M. R. et al. Pre-Clovis projectile points at the Debra L. Friedkin site, Texas– eaav2621 (2018).
implications for the Late Pleistocene peopling of the Americas. Sci. Adv. 4, eaat4505 46. Moreno-Mayar, J. V. et al. Terminal Pleistocene Alaskan genome reveals first founding
(2018). population of Native Americans. Nature 553, 203–207 (2018).
9. Haynes, G. The millennium before Clovis. PaleoAmerica 1, 134–162 (2015). 47. Ardelean, C. F. et al. Evidence of human occupation in Mexico around the Last Glacial
10. Sandweiss, D. H. et al. Quebrada Jaguay: early South American maritime adaptations. Maximum. Nature https://doi.org/10.1038/s41586-020-2509-0 (2020).
Science 281, 1830–1832 (1998). 48. Goodyear, A. C. in Paleoamerican Origins: Beyond Clovis (eds. Bonnichesen, R. et al.)
11. Goebel, T. & Keene, J. L. in Archaeology in the Great Basin and Southwest: Papers in 103–112 (Centre for the Study of the First Americans, 2005).
Honor of Don D. Fowler (eds Parezo, N. J. & Janetski, J. C.) 35–60 (Univ. Utah Press, 49. Llamas, B. et al. Ancient mitochondrial DNA provides high-resolution time scale of the
2014). peopling of the Americas. Sci. Adv. 2, e1501385 (2016).
12. Méndez, C., Jackson, D., Seguel, R. & Delaunay, A. N. Early high-quality lithic procurement 50. Pinotti, T. et al. Y chromosome sequences reveal a short Beringian standstill, rapid
in the semiarid north of Chile. Curr. Res. Pleistocene 27, 19–21 (2010). expansion, and early population structure of Native American founders. Curr. Biol. 29,
13. Méndez, C. & Jackson, D. Terminal Pleistocene lithic technology and use of space in 149–157.e3 (2019).
Central Chile. Chungara (Arica) 47, 53–65 (2015). 51. Bergström, A. et al. Insights into human genetic variation and population history from 929
14. Jones, K. B., Hodgins, G. W. L. & Sandweiss, D. H. Radiocarbon chronometry of site QJ- diverse genomes. Science 367, eaay5012 (2020).
280, Quebrada Jaguay, a Terminal Pleistocene to Early Holocene fishing site in southern 52. Raghavan, M. et al. Genomic evidence for the Pleistocene and recent population history
Peru. J. Island Coast. Archaeol. 14, 82–100 (2017). of Native Americans. Science 349, aab3884 (2015).

4 | Nature | www.nature.com
53. Rasmussen, M. et al. The genome of a Late Pleistocene human from a Clovis burial site in 68. Prescott, G. W., Williams, D. R., Balmford, A., Green, R. E. & Manica, A. Quantitative global
western Montana. Nature 506, 225–229 (2014). analysis of the role of climate and people in explaining Late Quaternary megafaunal
54. Reich, D. et al. Reconstructing Native American population history. Nature 488, 370–374 extinctions. Proc. Natl Acad. Sci. USA 109, 4527–4531 (2012).
(2012). 69. Cooper, A. et al. Abrupt warming events drove Late Pleistocene Holarctic megafaunal
55. Scheib, C. L. et al. Ancient human parallel lineages within North America contributed to a turnover. Science 349, 602–606 (2015).
coastal expansion. Science 360, 1024–1027 (2018). 70. Araujo, B. B. A., Oliveira-Santos, L. G. R., Lima-Ribeiro, M. S., Diniz-Filho, J. A. F. &
56. Erlandson, J. M. & Braje, T. J. From Asia to the Americas by boat? Paleogeography, Fernandez, F. A. S. Bigger kill than chill: the uneven roles of humans and climate on Late
paleoecology, and stemmed points of the northwest Pacific. Quat. Int. 239, 28–37 (2011). Quaternary megafaunal extinctions. Quat. Int. 431, 216–222 (2017).
57. Williams, T. J. & Madsen, D. B. The Upper Paleolithic of the Americas. PaleoAmerica 6, 71. Firestone, R. B. et al. Evidence for an extraterrestrial impact 12,900 years ago that
4–22 (2019). contributed to the megafaunal extinctions and the Younger Dryas cooling. Proc. Natl
58. Gilbert, M. T. P. et al. DNA from pre-Clovis human coprolites in Oregon, North America. Acad. Sci. USA 104, 16016–16021 (2007).
Science 320, 786–789 (2008). 72. Broughton, J. M. & Weitzel, E. M. Population reconstructions for humans and megafauna
59. Pedersen, M. W. et al. Postglacial viability and colonization in North America’s ice-free suggest mixed causes for North American Pleistocene extinctions. Nat. Commun. 9, 5441
corridor. Nature 537, 45–49 (2016). (2018).
60. Heintzman, P. D. et al. Bison phylogeography constrains dispersal and viability of the ice 73. Grayson, D. K. & Meltzer, D. J. Revisiting Paleoindian exploitation of extinct North
free corridor in western Canada. Proc. Natl Acad. Sci. USA 113, 8057–8063 (2016). American mammals. J. Archaeol. Sci. 56, 177–193 (2015).
61. Darvill, C. M., Menounos, B., Goehring, B. M., Lian, O. B. & Caffee, M. W. Retreat of the 74. Buck, C. E. & Bard, E. A calendar chronology for Pleistocene mammoth and horse
western Cordilleran Ice Sheet margin during the last deglaciation. Geophys. Res. Lett. 45, extinction in North America based on Bayesian radiocarbon calibration. Quat. Sci. Rev.
9710–9720 (2018). 26, 2031–2035 (2007).
62. Taylor, M. A., Hendy, I. L. & Pak, D. K. Deglacial ocean warming and marine margin retreat 75. Slon, V. et al. Neandertal and Denisovan DNA from Pleistocene sediments. Science 356,
of the Cordilleran Ice Sheet in the North Pacific Ocean. Earth Planet. Sci. Lett. 403, 89–98 605–608 (2017).
(2014). 76. Bueno, L., Politis, G., Prates, L. & Steele, J. A Late Pleistocene/early Holocene
63. Martin, P. S. The discovery of America. Science 179, 969–974 (1973). archaeological 14C database for Central and South America: palaeoenvironmental
64. Surovell, T. A., Pelton, S. R., Anderson-Sprecher, R. & Myers, A. D. Test of Martin’s overkill contexts and demographic interpretations. Quat. Int. 301, 1–158 (2013).
hypothesis using radiocarbon dates on extinct megafauna. Proc. Natl Acad. Sci. USA 113, 77. Bond, J. D. Paleodrainage map of Beringia, Yukon Geological Survey, open file 2019-
886–891 (2016). 2. http://data.geology.gov.yk.ca/Reference/81642#InfoTab (2019).
65. Robinson, G. S., Pigott Burney, L. & Burney, D. A. Landscape paleoecology and 78. Clark, G. A. in The Settlement of the American Continents (ed. Barton, C. M. et al.) 103–112
megafaunal extinction in southwestern New York state. Ecol. Monogr. 75, 295–315 (Univ. Arizona Press, 2004).
(2005).
66. Guthrie, R. D. New carbon dates link climatic change with human colonization and Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
Pleistocene extinctions. Nature 441, 207–209 (2006). published maps and institutional affiliations.
67. Lorenzen, E. D. et al. Species-specific responses of Late Quaternary megafauna to
climate and humans. Nature 479, 359–364 (2011). © The Author(s), under exclusive licence to Springer Nature Limited 2020

Nature | www.nature.com | 5
Article
Methods KDE_Plot can be used to estimate the underlying distribution of datasets
using kernel density estimation (KDE)84–86. KDE_Model, in particular, is
No statistical methods were used to predetermine sample size. The an MCMC approach that can be applied to sets of related dates. This uses
experiments were not randomized and investigators were not blinded a normal kernel, assumes a normal distribution (estimating bandwidth
to allocation during experiments and outcome assessment. according to Silverman’s rule87), applies a shaping parameter to the
Previously published archaeological and chronometric data for bandwidth to overcome over-smoothing of multi-modal distributions,
42 archaeological sites (Supplementary Information, Supplementary and treats events within a grouping as dependent. Default kernel and
Table 1) were used to construct site- and tradition-level models using factor parameters applied relative to Siverman’s rule in OxCal are N(0,1)
OxCal 4.3, a Bayesian modelling program19, and the IntCal13 calibration and U(0,1), accordingly.
curve21. In OxCal, models are built using chronological query language Further details, including mathematical notations, can be found
(CQL), which is identified here in italics. online in the OxCal manual88.

Bayesian age modelling and non-Markov chain Monte Carlo Model construction and visualization
functions in OxCal Archaeological sites with dated stratified sequences were incorporated
Bayesian age modelling is a statistical tool that enables the analysis of into single or multiple uniform phases, and start boundaries were used
chronometric data (standardized likelihoods) in the context of exist- as temporal markers of human presence (Supplementary Information).
ing, pertinent knowledge (priors) and the results obtained (posterior). Unstratified sites containing roughly synchronous archaeological
Priors can include absolute (for example, historical accounts) or rela- events—directly dated human burials with associated (dated) artefacts
tive information (for example, stratigraphy)79. Results are expressed or faunal remains subjected to human predation—were included in a
as probability density functions (PDFs) through the use of a random single, uniform Phase, and a Date function was used to provide an age
sampling technique, Markov chain Monte Carlo (MCMC), that uses a estimate. For these, if applicable and available, only the latest bone
mixture of the Metropolis–Hastings algorithm and Gibbs sampling80,81. collagen dates obtained through XAD89,90 or HYP91 protocols were
The program is set to initially analyse 30,000 iterations, with conver- used—recent studies show these to provide more reliable measure-
gence checked every 3,000 (a ‘pass’) and new starting points at every ments, particularly for contaminated bone5,92–94. If these sites were best
6,000. If convergence is less than 95%, the pass interval is increased by or solely represented by a single measurement, a calibrated date was
a factor of two until convergence is satisfactory82. As preset, PDFs are used. All modelled dates were assigned a 0.05 probability of being an
quoted as age ranges including the 68.2, 95.4 or 99.7% most likely results outlier (unless otherwise specified), and General and SSimple models
based on the prior (the highest posterior density range). In plots, these were used (using default parameters). We relied on outlier analysis to
are identifiable over individual calibrated radiocarbon likelihoods or objectively identify and downweight outliers within the models. Fol-
non-radiocarbon measurements as a darker distribution. Bayesian lowing these methods, site-specific age estimates (start boundaries,
age modelling assumes that time can be divided into units and treats Date functions or calibrated dates) were obtained for all archaeo-
archaeological events as groups defined by start and end boundaries. logical sites (Supplementary Information, Supplementary Table 2).
In this sense, a group can be incorporated into OxCal as a Phase within Where appropriate, these were incorporated into uniform phases
a Sequence constrained by two Boundary commands. A Phase contains to build tradition-level models and PDFs for the start of Beringian,
a group with unrelated, unordered items, whereas those in a Sequence Western Stemmed and Clovis traditions were obtained (Extended
follow an order. The type of group can be defined by the type of bound- Data Figs. 1–3). This allowed for the comparison of different tradi-
ary used (Boundary, Zero_Boundary, Tau_Boundary or Sigma_Boundary) tions spatiotemporally, as informed by the statistical assessment of
and the distribution each denotes—uniform, ramped, exponential or individual archaeological sequences. To visualize the temporal and/or
normal (figure 4 in ref. 19). A Boundary at the start and end of a group, spatial distribution of chronometric data, KDE methods were applied
for example, defines a uniform phase and assumes, a priori, that all using the KDE_Model function (using default parameters noted above)
events included are equally likely to occur anywhere within the two (Figs. 2, 3, Extended Data Fig. 4) and spatial KDE analysis within the
boundaries. An outlier analysis approach (Outlier_Model() and Outlier()) mapping package (see ref. 95 for a description of the algorithm) to
can be used to address the assumption that all dates are correct. This identify clusters within the model output (Extended Data Fig. 4). All
defines the distribution and scale of outliers, and assigns each date a age estimates are noted to 95.4% confidence or credible (for Bayesian
prior probability of being an outlier. The General outlier model, for model output) interval, unless otherwise specified, and rounded to
example, draws outliers from a Student’s t-distribution with 5 degrees five years (following a convention explained in ref. 96).
of freedom to account for extreme outliers, and sets the scale of offset
to range widely, between 100 to 104 years83. The analysis then identi- Reporting summary
fies outliers by assessing the fit of chronometric data within the prior Further information on research design is available in the Nature
framework, and downweights them automatically according to their Research Reporting Summary linked to this paper.
degree of offset. This approach is more objective than manual exclu-
sion of dates and has a similar output (figure 9 in ref. 83). In archaeology,
outlier analysis can be used to assess agreement between chronometric Data availability
data and stratigraphic evidence, often yielding important information The data that support the findings of this study are available in the
regarding site-specific processes (for example, geological unconform- Article and its Supplementary Information.
ity or bioturbation).
In OxCal, there are a number of MCMC and non-MCMC functions that
Code availability
are relevant to archaeological investigations. These can be used to pro-
duce age estimates for archaeological features (Date within a Phase; this Code for OxCal is noted in the Supplementary Information.
produces a distribution for the dated events) or the commencement of
79. Harris, E. C. Principles of Archaeological Stratigraphy (Elsevier, 2014).
an occupational phase (start Boundary), or to test the goodness of fit 80. Gelfand, A. E. & Smith, A. F. M. Sampling-based approaches to calculating marginal
between multiple dates relating to the same moment in time (R_Com- densities. J. Am. Stat. Assoc. 85, 398–409 (1990).
bine and Combine). Modelling the start of an archaeological occupa- 81. Gilks, W. R., Richardson, S. & Spiegelhalter, D. Markov Chain Monte Carlo in Practice
(Chapman and Hall/CRC, 1995).
tion is particularly useful in the study of human dispersals, as human 82. Bronk Ramsey, C. Radiocarbon calibration and analysis of stratigraphy: the OxCal
presence precedes an archaeological event. In addition, KDE_Model and program. Radiocarbon 37, 425–430 (1995).
83. Bronk Ramsey, C. Dealing with outliers and offsets in radiocarbon dating. Radiocarbon 51, 95. Bronk Ramsey, C., Housley, R. A., Lane, C. S., Smith, V. C. & Pollard, A. M.
1023–1045 (2009). The RESET tephra database and associated analytical tools. Quat. Sci. Rev. 118, 33–47
84. Rosenblatt, M. Remarks on some nonparametric estimates of a density function. Ann. (2015).
Math. Stat. 27, 832–837 (1956). 96. Derek Hamilton, W. & Krus, A. M. The myths and realities of Bayesian chronological
85. Parzen, E. On estimation of a probability density function and mode. Ann. Math. Stat. 33, modeling revealed. Am. Antiq. 83, 187–203 (2018).
1065–1076 (1962).
86. Bronk Ramsey, C. Methods for summarizing radiocarbon datasets. Radiocarbon 59, 1809–
1833 (2017). Acknowledgements Without implying their agreement with the content of this article,
87. Silverman, B. W. Density Estimation for Statistics and Data Analysis (Chapman & Hall, 1986). we thank E. Jacob, G. Wali, J. Swift, C. Bronk Ramsey, J. Lee-Thorp, K. Graf and K. Douka for
88. Bronk Ramsey, C. OxCal 4.3 Manual. https://c14.arch.ox.ac.uk/oxcalhelp/hlp_contents. their feedback on versions of the manuscript. We are grateful to the staff of the Oxford
html (2020). Radiocarbon Accelerator Unit, University of Oxford. Funding was provided by the
89. Stafford, T. W. Jr, Duhamel, R. C., Haynes, C. V. Jr & Brendel, K. Isolation of proline and Clarendon Fund Scholarship, University of Oxford.
hydroxyproline from fossil bone. Life Sci. 31, 931–938 (1982).
90. Stafford, T. W., Brendel, K. & Duhamel, R. C. Radiocarbon, 13C and 15N analysis of fossil bone: Author contributions L.B.-V. compiled archaeological and chronometric data and built
removal of humates with XAD-2 resin. Geochim. Cosmochim. Acta 52, 2257–2267 (1988). Bayesian age models. L.B.-V. and T.H. analysed modelled output and wrote the manuscript.
91. Deviese, T., Comeskey, D., McCullagh, J., Bronk Ramsey, C. & Higham, T. New protocol for
compound-specific radiocarbon analysis of archaeological bones. Rapid Commun. Mass Competing interests The authors declare no competing interests.
Spectrom. 32, 373–379 (2018).
92. Devièse, T. et al. Increasing accuracy for the radiocarbon dating of sites occupied by the Additional information
first Americans. Quat. Sci. Res. 198, 171–180 (2018). Supplementary information is available for this paper at https://doi.org/10.1038/s41586-020-
93. Becerra-Valdivia, L. et al. Reassessing the chronology of the archaeological site of Anzick. 2491-6.
Proc. Natl Acad. Sci. USA 115, 7000–7003 (2018). Correspondence and requests for materials should be addressed to L.B.-V.
94. Waters, M. R., Stafford, T. W. Jr, Kooyman, B. & Hills, L. V. Late Pleistocene horse and camel Peer review information Nature thanks Loren G. Davis, Christopher L. Hill and the other,
hunting at the southern margin of the ice-free corridor: reassessing the age of Wally’s anonymous, reviewer(s) for their contribution to the peer review of this work.
Beach, Canada. Proc. Natl Acad. Sci. USA 112, 4263–4267 (2015). Reprints and permissions information is available at http://www.nature.com/reprints.
Article

Extended Data Fig. 1 | Bayesian age model and start boundary for the Beringian tradition (14,955–13,895 cal. bp). Right, estimate rounded to 50. Outlier
analysis output (O) is noted as ‘posterior probability/prior probability’. δ18O data are according to the Greenland ice-core timescale (GICC05)20.
Extended Data Fig. 2 | Bayesian age model and start boundary for the Western Stemmed tradition (14,860–13,065 cal. bp). Outlier analysis output (O) is
noted as ‘posterior probability/prior probability’. δ18O data are according to the Greenland ice-core timescale (GICC05)20.
Article

Extended Data Fig. 3 | Bayesian age model and start boundary for Clovis tradition (14,210–13,495 cal. bp). Outlier analysis output (O) is noted as ‘posterior
probability/prior probability’. δ18O data are according to the Greenland ice core timescale (GICC05)20.
Extended Data Fig. 4 | Spatio-temporal slices of chronometric data (Methods). For each date, differences in circle size reflect increasing or
belonging to the cultural components analysed, with a spatial KDE decreasing probabilities at a 95.4% confidence interval. The spatial KDE
analysis. a–f, Coloured circles (following colour scheme in Fig. 1) denote analysis shows a marked increase in the frequency and distribution of the data
chronometric data (n = 387 dates) and white outlines reflect the spatial KDE immediately, before and during GI-1.
analysis. Chronometric data were summarized using a KDE_Model analysis
 nature research | reporting summary
Corresponding author(s): Lorena Becerra-Valdivia
Last updated by author(s): May 13, 2020

Reporting Summary
Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency
in reporting. For further information on Nature Research policies, see Authors & Referees and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.
n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement
A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly
The statistical test(s) used AND whether they are one- or two-sided
Only common tests should be described solely by name; describe more complex techniques in the Methods section.

A description of all covariates tested

A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons
A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient)
AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)

For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
Give P values as exact values whenever suitable.

For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes
Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated
Our web collection on statistics for biologists contains articles on many of the points above.

Software and code

Policy information about availability of computer code
Data collection Archaeological and chronometric data were collated by the first author from previously published sources (as detailed in the
Supplementary Information), in the software listed below and institutional computers/databases, following the appropriate guidelines.

Data analysis This project used OxCal, version 4.3 (2020), which is freely available online.
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors/reviewers.
We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.

Data
Policy information about availability of data
All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:
- Accession codes, unique identifiers, or web links for publicly available datasets
- A list of figures that have associated raw data
- A description of any restrictions on data availability

Radiocarbon and OSL measurements, associated archaeological information, calibrated date ranges and CQL codes for Bayesian models are included in the
October 2018

Supplementary Information.

1
 nature research | reporting summary
Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.
Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Ecological, evolutionary & environmental sciences study design

All studies must disclose on these points even when the disclosure is negative.
Study description Bayesian age modeling of archaeological and chronometric data pertaining to the initial peopling of the Americas, specifically North
America and Beringia.

Research sample Archaeological and chronometric data pertaining to the initial peopling of the Americas, specifically North America and Beringia

Sampling strategy N=767 for chronometric data and N=42 for archaeological sites included. All data are duly described in the Supplementary
Information.

Data collection The first author, Dr Becerra-Valdivia, collated the chronometric and archaeological data into spreadsheets archived at her institution
according to the appropriate archival practices.

Timing and spatial scale Data was collected between 2016 and 2019 at the University of Oxford, United Kingdom.

Data exclusions No data pertaining to the cultural components studied were excluded for single site models. Archaeological sites that were excluded
from the analysis are described in the Supplementary Information, along with the reasoning for their exclusion.

Reproducibility Sensitivity testing for Bayesian age modeling was carried out and described in the Supplementary Information.

Randomization Randomization is not relevant to this study as there is no experiment being conducted. We report on direct chronometric and
archaeological data.

Blinding Blinding is not relevant to this study as there is no experiment being conducted; the chronometric and archaeological data used is
freely available as previously published.

Did the study involve field work? Yes No

Reporting for specific materials, systems and methods

We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material,
system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.

Materials & experimental systems Methods

n/a Involved in the study n/a Involved in the study
Antibodies ChIP-seq
Eukaryotic cell lines Flow cytometry
Palaeontology MRI-based neuroimaging
Animals and other organisms
Human research participants
Clinical data
October 2018