Redefining the Age of Clovis: Implications for the Peopling of the Americas

See allHide authors and affiliations

Science  23 Feb 2007:
Vol. 315, Issue 5815, pp. 1122-1126
DOI: 10.1126/science.1137166


The Clovis complex is considered to be the oldest unequivocal evidence of humans in the Americas, dating between 11,500 and 10,900 radiocarbon years before the present (14C yr B.P.). Adjusted 14C dates and a reevaluation of the existing Clovis date record revise the Clovis time range to 11,050 to 10,800 14C yr B.P. In as few as 200 calendar years, Clovis technology originated and spread throughout North America. The revised age range for Clovis overlaps non-Clovis sites in North and South America. This and other evidence imply that humans already lived in the Americas before Clovis.

For nearly 50 years, it has been generally thought that small bands of humans carrying a generalized Upper Paleolithic tool kit entered the Americas around 11,500 radiocarbon years before the present (14C yr B.P.) and that these first immigrants traveled southward through the ice-free corridor separating the Laurentide and Cordilleran Ice Sheets (1). These people developed the distinctive lithic, bone, and ivory tools of Clovis (2, 3) and then quickly populated the contiguous United States. Clovis humans and their descendants then rapidly populated Central America and reached southernmost South America by 10,500 14C yr B.P. (1).

Identifying when the Clovis complex first appeared and knowing the complex's duration is critical to explaining the origin of Clovis, evaluating the Clovis-first model of colonization of the Americas, determining the role of humans in the extinction of late Pleistocene megafauna, and assessing whether people inhabited the Americas before Clovis. We determined a more accurate time span for Clovis by analyzing the revised existing Clovis 14C date record and reporting high-precision accelerator mass spectrometry (AMS) 14C ages from previously dated Clovis sites. Our AMS 14C dates are on culturally specific organic matter—bone, ivory, and seeds—that accelerator mass spectrometers can date accurately (4, 5) to precisions of ±30 years at 11,000 14C yr B.P.

Clovis technology has strong Old World antecedents, but Clovis-specific traits (e.g., fluted lanceolate projectile points) probably originated in the New World, south of the continental ice sheets (3). Clovis tools and debitage identify and unify archaeological sites over a broad geographic range. Clovis sites and artifacts cluster in North America, especially in the contiguous United States (1). A small number of Clovis artifacts have been recovered from Mexico and possibly as far south as Venezuela (6). Even though Clovis covers a broad geographic range, only 22 Clovis sites in North America have been directly 14C-dated (Fig. 1, Table 1, and table S1). The 14C dates from these sites traditionally place Clovis between 11,500 and 10,900 14C yr B.P. (1, 7, 8). However, the 14C dates from 11 of these sites are problematic and do not provide accurate or precise chronological information to determine the age of Clovis (5).

Fig. 1.

Map showing the location of Clovis and other early sites. The numbers correspond to those found in Table 1. Other sites are 31, Monte Verde, Chile; 32, Nenana Complex sites, Alaska; and 33, Broken Mammoth, Alaska.

Table 1.

Summary of 14C dates from Clovis and Clovis-age sites. Single 14C dates, date ranges, and averaged dates are reported. If multiple 14C dates were available from a single-component site, the dates were averaged with the method in (28). All dates are given at 1σ SD. n, number of dates.

SiteDate (14C yr B.P.)
Clovis sites (credible ages and Clovis diagnostics)
1. Lange-Ferguson, SD (n = 3) 11,080 ± 40
2. Sloth Hole, FL (n = 1) 11,050 ± 50
3. Anzick, MT (foreshaft ages) (n = 2) 11,040 ± 35
4. Dent, CO (n = 3) 10,990 ± 25
5. Paleo Crossing, OH (n = 3) 10,980 ± 75
6. Domebo, OK (n = 1) 10,960 ± 30
7. Lehner, AZ (n = 12) 10,950 ± 40
8. Shawnee-Minisink, PA (n = 5) 10,935 ± 15
9. Murray Springs, AZ (n = 8) 10,885 ± 50
10. Colby, WY (n = 2) 10,870 ± 20
11. Jake Bluff, OK (n = 3) 10,765 ± 25
Clovis sites (indirectly dated and Clovis diagnostics)
12. East Wenatchee, WA (n = 1) <11,125 ± 130
Clovis-age sites (credible ages but no Clovis diagnostics)
13. Indian Creek, MT (n = 1) 10,980 ± 110
14. Lubbock Lake, TX (n = 2) 11,100 ± 60
15. Bonneville Estates, NV (n = 1) 11,010 ± 40
16. Kanorado, KS (n = 2) 10,980 ± 40
17. Arlington Springs, CA (n = 1) 10,960 ± 80
Problematic Clovis and Clovis-age sites
18. Sheriden Cave, OH (above artifacts, n = 5) 10,600 ± 30
Sheriden Cave, OH (below artifacts, n = 2) 10,920 ± 50
19. Blackwater Draw, NM (n = 3) 11,300 ± 235
20. Cactus Hill, VA (n = 1) 10,920 ± 250
21. Wally's Beach, Canada (n = 4) 11,350 ± 80 to 10,980 ± 80
22. Union Pacific, WY (n = 1) 11,280 ± 350
23. Aubrey, TX (n = 2) 11,570 ± 70
24. Sheaman, WY (n = 3) 10,305 ± 15
Ages from other early sites
25. Mill Iron, MT (Goshen) (n = 4) 10,840 ± 60
26. Hell Gap, WY (Goshen) (n = 1) 10,955 ± 135
27. Cerro Tres Tetas, Argentina (pre-Fishtail, n = 5) 10,935 ± 35
28. Cuevas Casa del Minero, Argentina (pre-Fishtail, n = 2) 10,985 ± 40
29. Piedra Museo, Argentina (pre-Fishtail, n = 2) 10,960 ± 45
30. Fell's Cave, Chile (Fishtail, n = 1) 11,000 ± 170

Three sites (East Wenatchee, Washington; Blackwater Draw, New Mexico; and Cactus Hill, Virginia) have Clovis diagnostic artifacts but lack precise ages (5). Three sites (Lubbock Lake, Texas; Kanorado, Kansas; and Indian Creek, Montana) fall within the Clovis age range but lack diagnostic Clovis artifacts (5). The site of Sheriden Cave, Ohio, provides only bracketing ages for Clovis artifacts (5). Questions exist about the accuracy of the 14C dates from Aubrey, Texas (5), where diagnostic Clovis artifacts were found. We obtained three dates from the Sheaman site, Wyoming, that averaged 10,305 ± 15 14C yr B.P. These dates indicate that the Clovis context at Sheaman is mixed with younger cultural materials (5). Finally, associations between Clovis artifacts and 14C-dated faunal remains at two sites (Wally's Beach, Canada; and Union Pacific, Colorado) are unresolved (5). Because of these problems, we excluded the dates from these sites in assessing the age of Clovis.

This leaves 11 sites with a total of 43 14C dates (Table 1 and table S1) (5). These sites have assemblages of Clovis artifacts in secure geological contexts. Existing ages from five sites (Anzick, Montana; Paleo Crossing, Ohio; Lehner, Arizona; Murray Springs, Arizona; and Jake Bluff, Oklahoma) already have high-precision 14C dates on credible materials. We obtained nine new ages from seeds and highly purified bone and ivory collagen for five imprecisely dated sites (Lange-Ferguson, South Dakota; Dent, Colorado; Domebo, Oklahoma; Shawnee-Minisink, Pennsylvania; and Colby, Wyoming) (4, 5). In addition, we obtained five ages on human remains from the Anzick site, Montana (5). We attempted to date samples from Sloth Hole, Florida, but the samples contained no collagen.

These 43 14C dates place the beginning of Clovis at ∼11,050 14C yr B.P. (reducing former estimates by 450 14C years) and its end at ∼10,800 14C yr B.P. (younger than previous estimates by 100 14C years). Accurate calendar correlation of 14C ages from the Clovis time period is not currently possible because of correlation uncertainties (9). The Clovis-period segment of the INTCAL04 calibration is based on 14C-dated marine foraminifera and is not accurate for the Clovis time period (10). The most accurate calibration for this time period is provided by a floating European tree-ring chronology that is provisionally anchored to INTCAL04 (11). Using this tentative calibration (11), we estimated that Clovis has a maximum possible date range of 13,250 to 12,800 calendar yr B.P.—a span of 450 calendar years (Fig. 2). By taking the youngest possible calibrated age for the oldest Clovis site and the oldest possible calibrated age for the youngest Clovis site, a minimum range for Clovis is calculated as 13,125 to 12,925 calendar yr B.P.—a span of 200 calendar years. The ages for all Clovis sites overlap within this 200-year period, and this time span probably represents the true range of Clovis. However, the absolute calendar placement of the floating tree-ring record is disputed (12). By an alternative calibration (12), the maximum time range for Clovis is 13,110 to 12,660 calendar yr B.P., and the minimum time range is 12,920 to 12,760 calendar yr B.P. (Fig. 3). Regardless of the exact calendar dates, the 200-year duration for Clovis remains secure because the floating dendrochronological sequence provides calendar-year separations between two 14C-dated sites.

Fig. 2.

Calendar-year age ranges for Clovis and other early sites based on the European dendrochonological calibration (11) at 1σ SD.

Fig. 3.

Calendar-year age ranges for Clovis and other early sites based on the Fairbanks calibration (12) at 1σ SD.

The oldest Clovis sites (n = 3 sites) are located in Montana, South Dakota, and Florida; younger Clovis sites are located in the interior (n = 5) of the United States and in the South-west (n =2) and East (n = 1). The distribution of dated sites shows no clear indication of north-south or east-west age differences that would indicate movement of people in one direction or another. Instead, Clovis technology seems to have appeared synchronously across the United States at ∼11,050 14C yr B.P. This pattern of 14C dates is compatible with two contrasting hypotheses.

First, this pattern could support the idea that there was a rapid spread of Clovis people across an empty continent. Demographic models suggest that people exiting the ice-free corridor could have occupied the contiguous United States within 100 years or less (13). Although there is much speculation about a coastal migration of the first Americans from both Asia and Europe (14, 15), the revised date range for Clovis reopens the possibility of a Late Glacial migration through the ice-free corridor that separated the Laurentide and Cordilleran Ice Sheets. People could have easily traveled through the ice-free corridor after ∼11,500 14C yr B.P. (1)—at least 200 calendar years before the oldest known Clovis date. The biface and blade industry of Nenana (16) was well established at the Broken Mammoth site, Alaska, to 11,770 ± 210 14C yr B.P. (WSU-4351)—at least 300 calendar years before our oldest recalibrated Clovis date. The Nenana lithic assemblage shows strong similarities to the Clovis lithic assemblage (17). It is possible that either Nenana people or others with a biface and blade industry traveled through the corridor, and once south of the ice sheets, they developed the technological hallmarks characteristic of Clovis and spread rapidly across the continent.

An alternative interpretation is that the instantaneous appearance of Clovis across North America represents the rapid spread of Clovis technology through a preexisting but culturally and genetically undefined human population in North America (18). In this case, Clovis technology could have been introduced to this population through a Late Glacial migration of Clovis or Clovis progenitors or developed in situ from a pre-Clovis technology already in the Americas. Regardless of which hypothesis is correct, our revised chronology indicates that Clovis technology spread rapidly.

Faunal remains associated with dated Clovis sites constrain the timing of the extinction of Proboscideans at the end of the Pleistocene. Mammoths and mastodons were an important source of food and raw materials used to manufacture bone and ivory tools (3), as well as perishable items from soft tissues. Proboscidean remains are associated with seven of the well-dated Clovis sites (Lange-Ferguson, Sloth Hole, Dent, Domebo, Lehner, Murray Springs, and Colby), and the last occurrence of mammoth in the United States is dated at ∼10,900 14C yr B.P. After this time, Clovis and sites of other complexes (e.g., Goshen and Folsom) contained only bison and other extant species. The extinction of mammoth and mastodon coincides with the main florescence of Clovis.

Our revised ages for Clovis overlap dates from a number of North American sites that are technologically or culturally not Clovis. The earliest dated sites of the Goshen complex (Mill Iron, Montana; and Hell Gap, Wyoming) (19) overlap the age range of Clovis (Figs. 2 and 3, and Table 1, and table S1). This indicates that Goshen is either coeval with the entire range of Clovis or briefly overlaps the later stages of the Clovis time period. Clovis also overlaps the date for the Arlington Springs human skeleton from Santa Rosa Island, California (Figs. 2 and 3 and Table 1) (20). No artifacts were found with the Arlington Springs human remains, and his cultural affiliation is unknown. The presence of human remains on Santa Rosa Island is unequivocal evidence that water crafts were used during Clovis time and that a Pacific maritime-coastal adaptation was probably contemporaneous with Clovis. At Bonneville Estates Rockshelter, Nevada (21), the earliest date from a series of hearths is coeval with the Clovis time period and is associated with stone artifacts. Diagnostic artifacts have yet to be found at this level, and it is unknown whether this early horizon is associated with Clovis or stemmed points.

Several sites in South America have yielded 14C dates that are coeval with Clovis (Figs. 2 and 3). These include the early archaeological horizons at Cerro Tres Tetas, Cueva Casa del Minero, and Piedra Museo, Argentina, and the earliest Fishtail point horizon at Fell's Cave, Chile (Fig. 1, Table 1, and table S1) (8, 22). The actual calendar dates of these South American sites may be slightly more recent because Southern Hemisphere samples have a lower initial 14C content than contemporaneous samples in the Northern Hemisphere. This latitudinal difference causes Southern Hemisphere terrestrial materials to be 5 to 80 14C years older than contemporaneous samples in North America (23). The extent of this interhemisphere offset in 14C years for the Clovis time period is unknown, but it is probably less than 80 years. Even with an 80-year correction to the 14C dates from these four sites, Clovis, Fishtail, and other early complexes in the Southern Cone of South America are still contemporaneous.

The presence of non-Clovis sites that are contemporaneous with Clovis in both North and South America implies that Clovis does not represent the earliest occupation of the Americas. It would probably have taken a minimum of 600 to 1000 years for the first Paleoamericans and their descendents to travel by land from the southern limit of the ice-free corridor to Tierra del Fuego (13, 24)—a distance of over 14,000 km (Fig. 1). However, at most 300 to 350 calendar years separate the oldest possible date for Clovis and the youngest possible ages from the well-dated unequivocal sites in southernmost South America (Figs. 2 and 3). The difference is reduced to 200 calendar years, when the minimum date for the beginning of Clovis in North America and the youngest dates for the sites in South America are considered. It is highly improbable that within 200 to 350 calendar years, people entered North America; adapted to biomes ranging from artic tundra to grasslands, deserts, and rainforests; increased in population; and reached the southern tip of South America within the span of 10 to 18 human generations. This suggests that human populations already existed in the New World before Clovis.

There is an emerging archaeological record that supports a pre-Clovis human occupation of the Americas. Stone tools and butchered mammoth remains dating to ∼12,500 14C yr B.P. have been found at the Schaefer and Hebior sites in Wisconsin (25, 26). Older butchered mammoth remains dating to ∼13,500 14C yr B.P. have been recovered from the Mud Lake site, Wisconsin (25, 26). In South America, humans appear to have been present at 12,500 14C yr B.P. at Monte Verde, Chile (27). The archaeological data now show that Clovis does not represent the earliest inhabitants of the Americas and that a new model is needed to explain the peopling of the Americas.

Supporting Online Material

Materials and Methods

SOM Text

Table S1


References and Notes

View Abstract

Navigate This Article