Letters to Nature

Nature 429, 746-749 (17 June 2004) | doi:10.1038/nature02612; Received 8 March 2004; Accepted 27 April 2004

Radiocarbon evidence of mid-Holocene mammoths stranded on an Alaskan Bering Sea island

R. Dale Guthrie1

  1. Institute of Arctic Biology, University of Alaska, Fairbanks, Alaska 99775, USA

Correspondence to: R. Dale Guthrie1 Email: ffrdg@uaf.edu


Island colonization and subsequent dwarfing of Pleistocene proboscideans is one of the more dramatic evolutionary and ecological occurrences1, 2, 3, especially in situations where island populations survived end-Pleistocene extinctions whereas those on the nearby mainland did not4. For example, Holocene mammoths have been dated from Wrangel Island in northern Russia4. In most of these cases, few details are available about the dynamics of how island colonization and extinction occurred. As part of a large radiocarbon dating project of Alaskan mammoth fossils, I addressed this question by including mammoth specimens from Bering Sea islands known to have formed during the end-Pleistocene sea transgression5. One date of 7,908 plusminus 100 yr bp (radiocarbon years before present) established the presence of Holocene mammoths on St Paul Island, a first Holocene island record for the Americas. Four lines of evidence—265 accelerator mass spectrometer (AMS) radiocarbon dates from Alaskan mainland mammoths6, 13 new dates from Alaskan island mammoths, recent reconstructions of bathymetric plots5 and sea transgression rates from the Bering Sea5—made it possible to reconstruct how mammoths became stranded in the Pribilofs and why this apparently did not happen on other Alaskan Bering Sea islands.

During the first stages of a Holocene sea level rise at 13,000 radiocarbon years before present (referred to hereafter as yr bp), the Pribilofs were simply uplands on a large island lying within a context of dozens of small islands (Fig. 1b). By the time of the St Paul mammoth (Mammuthus primigenius) date, 7,900 yr bp (22 m below current sea level), inundation had claimed more land and the uplands that were to become the Pribilofs were separated into several smaller islands quite distant from the mainland (Fig. 1e). Even at that relatively high mid-Holocene sea stand, St Paul Island was approximately 5–10 times larger than today. Some time after 5,000 yr bp (4 m below current sea level) St Paul Island was reduced to near its present size (91 km2) and may have ultimately been too small (Fig. 1f) to support populations of mammoth. Small islands have a limited carrying capacity, and at low numbers inbreeding depression becomes debilitating, then lethal7, 8. The absolute size at which an island could no longer support a proboscidean population would depend on productivity of appropriate forage, but judging from empirical evidence the minimum may be over 100 km2.

Figure 1: Rise in sea level of the Bering Sea with time, modified from Manley's approximate contours and transgression dates10.
Figure 1 : Rise in sea level of the Bering Sea with time, modified from Manley's approximate contours and transgression dates10. Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, or to obtain a text description, please contact npg@nature.com

af, Mammoth distribution (red) at 18,000 (a), 13,000 (b), 10,000 (c), 9,000 (d), 8,000 (e) and 4,000 (f) yr bp is shown. This reconstruction shows how mammoths would have been isolated on the large island containing the Pribilofs around 13,000 yr bp. By 8,000 yr bp the Pribilofs were further isolated but were much larger in size than today. Note how St Lawrence and St Matthew were still extensions of the mainland until around 10,000 yr bp, well after woolly mammoths are suspected to have become extinct on the AK–YT mainland (see Supplementary Information).

High resolution image and legend (80K)

The date obtained from a small, upper third molar (M3) of the St Paul Island mammoth differs markedly in comparison with the date from the last Alaskan mainland mammoth fossil (11,500 plusminus 150 yr bp; Fig. 2). At that time (and perhaps even as recently as 10,000 yr bp) St Lawrence, St Matthew, Nunivak and Unimak islands were still connected to the mainland (Fig. 1c). But as the sea began to rise the development of Bering Sea islands was not synchronous. We know from dated mammoth bones found on St Lawrence Island that these unglaciated9 uplands were used by mammoths during the Last Glacial Maximum (LGM), around 18,000 yr bp when the sea level was 120 m below its current level, but none of the fossils had post-Pleistocene dates (see Supplementary Information). Whatever forces caused mainland mammoth extinction in Alaska and the Yukon Territory (AK–YT) were probably also present in regions that would become islands after 11,500 yr bp. Thus, there is little reason to expect to find Holocene mammoths from St Lawrence and other Bering Sea islands that became isolated after extinction of mammoths from the mainland. In contrast, the Pribilofs were separated from the mainland possibly as early as 13,000 yr bp (88 m below current sea level), when woolly mammoth were still widespread6, 10. In addition to mammoth, caribou (Rangifer), bear (Ursus) and fox (Alopex) fossils have been found on St Paul Island, and are currently under study11.

Figure 2: Rank comparison by AMS radiocarbon dates of woolly mammoth from the Alaskan and northwest Canadian mainland showing their relation to the specimen from the Pribilofs.
Figure 2 : Rank comparison by AMS radiocarbon dates of woolly mammoth from the Alaskan and northwest Canadian mainland showing their relation to the specimen from the Pribilofs. Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, or to obtain a text description, please contact npg@nature.com

Error bars = 1sigma (see Supplementary Information). The drawing is of the Pribilof mammoth tooth, with a dashed-line reconstruction of its original contours. Infinite dates beyond 40,500 yr bp are deleted here, but are listed in the Supplementary Information.

High resolution image and legend (36K)

Sea transgression reconstruction (Fig. 1c, d) makes it clear that other islands existed during the Early Holocene but were eventually inundated. These unnamed islands may have also sustained stranded mammoth populations. Across the Bering Sea, radiocarbon dates show that mammoth populations lasted until after 10,000 yr bp in northernmost Siberia10. This latter date is germane to the issue of why there is a gap in mammoth dates on Wrangel Island between 12,000 and 8,000 yr bp. Wrangel Island was part of the mainland around 12,000 yr bp but was many kilometres offshore at 8,000 yr bp. Did mammoths somehow recolonize Wrangel Island at 8,000 yr bp or is this gap a taphonomic artefact of what was actually continuous occupation. I leave a question mark (?) on that issue in Fig. 1c.

The productivity on St Paul Island today is rather modest12 compared with nearby St George Island and the Aleutians. Whereas the latter two are coastal grasslands, St Paul includes some tundra elements12, 13. Geological evidence supports the idea that during the Pleistocene this region was quite arid, experiencing loess deposition and reworking. Bathymetric analysis shows that the Yukon River emptied into the Bering Sea just to the south of the Pribilofs during the LGM. Evidence from a pollen core13 on Cagalog Lake on St Paul Island shows a dominance of grasses, sedges and Artemisia (sage), the ubiquitous spectra typical of cold/arid steppes of the unglaciated north in pre-Holocene times14, 15. However, during the Early Holocene (exact dates are disturbed by groundwater movement) on St Paul, the Umbelliferae forbs (probably Coeloleurum gmelini) increased in number, suggesting more moisture, a characteristic of Bering Sea rim vegetation today12, 13. St Paul Island does not have permafrost, and the season of snow cover is much shorter than its latitudinal equivalent on the mainland. The maritime buffer of the nearby north Pacific and accompanying weather tracts bring unbroken year-round low cloud cover, producing a cool, moist summer with a long growing season in which certain forbs flourish but trees are excluded.

There are two issues to address: why did mammoths survive later on the islands (St Paul, Wrangel and perhaps other, now inundated, islands) than on the mainland, and why did they fail to survive past a certain point. Proponents of the human overkill theory16 argue that whereas humans were responsible for the extinction of mammoths on the mainland, offshore island populations were spared because humans had limited access to distant islands until later in the Holocene. Those that see the Late Pleistocene extinctions as the result of ecological changes17 argue that island mammoths would have been sheltered from the profound climatic and vegetational shifts that favoured the explosion of Holocene large mammal species in competition with Pleistocene ones. However, that controversy remains unresolved.

It is possible that the extinction of island proboscideans may have been a different process than extinction on the mainland. Extinction of mammoths from the California Channel Islands is correlated with the documented arrival of humans around 11,000 yr bp18. The time of the first human occupation of Wrangel Island is uncertain, but the late timing of mammoth extinction there at around 3,700 yr bp does coincide with the time of marine expansion of northern Denbigh-Arctic small-tool-tradition peoples at about 4,000 yr bp19. There is archaeological evidence of human occupation of Wrangel Island at least at about 4,300 yr bp20. There is, however, no reason to suspect that humans killed mammoths on St Paul Island; we have no record of human discovery of that island until modern times. Rather, it appears that on St Paul sea transgression separated mammoths from the mainland and a continuation of that sea transgression eventually reduced island areas to the point where life was unsustainable for mammoths.

Closely linked to this matter of Pleistocene island extinction is the controversy over the timing and routes by which humans first entered North America. A 'coastal route' camp21 argues for an early colonization (around 13,000 yr bp), whereas the 'overland' camp18 points to a route between the Cordilleran–Laurentide glaciers, and highlights evidence in the range of 12,000–11,000 yr bp. The St Paul Island mammoth brings a new piece to this puzzle.

If there were coastal watercraft colonists, and they became the continent's 'super-hunters' as per the overkill theory, we might ask why did they not find and kill off mammoths on St Paul? Mammoths on that earlier island complex at 13,000 yr bp would have been easily visible in a treeless landscape when St Paul was separated from the mainland by a narrow channel, in the most likely path of coastal watercraft colonists (Fig. 1b). This becomes part of a broader issue because dwarf mammoths of the California near-shore Channel Islands pose a similar situation. There is no evidence that the Channel Island dwarf mammoths were hunted at around 13,000 yr bp, and they did not become extinct at that time. Rather, these island mammoths became extinct at the time Clovis-aged people invaded the islands at around 11,000 yr bp21. Clearly, some island mammoth extinctions were the result of human colonization; however, on St Paul that does not seem to have been the case.



Despite ground surveys and excavations on St Matthew Island for potential Quaternary outcrops, and excavated test pits, other researchers and myself have failed to find evidence of Quaternary fossils there. I searched the large collections from Alaska in the University of Alaska Museum, Fairbanks, US National Museum, Washington, and American Museum of Natural History, New York, for fossil bones from the Bering Sea islands. Those specimens are listed in the Supplementary Information. All specimens in this study were dated by the AMS radiocarbon-dating method at the NSF-University of Arizona AMS Facility at Tucson.

The Pribilof M3 was located at the US National Museum (USNM 23455). Provenance data indicate that the tooth was excavated by R. E. Carroll in 1964 from Northeast Point, 75 cm below the surface. The M3 showed a mid-range stage of wear, indicating an adult animal. Maximum width at the 12th plate was 6.5 cm, frequency of enamel plates was 9.23 per 10 cm, and the enamel thickness averaged around 1.2 mm. The roots had been abraded away almost to the ventral bases of the enamel plates, resulting in a smoothly worn contour, suggesting gentle rounding forces like wind-blown sand. The incompleteness of the St Paul specimen makes it difficult to judge body size. Island dwarfing is a general rule for large mammals2. Such dwarfing is a process of genetic change, but proboscideans are also known to possess unusual developmental size plasticity in response to conditions during their life22, 23. Ongoing work on St Paul may furnish an answer to this question of dwarfing.

I double-checked the first date of 7,908 plusminus 100 yr bp from the NSF-Arizona Laboratory (AA26010). That split-sample (AA34501) dated at 8,015 plusminus 85 yr bp. The sample was further re-dated at the Oxford ORAU facility at 8,010 plusminus 40 yr bp (OxA-13027).

The stable isotope 13C-21.5 analysis from the University of Arizona laboratory for the Pribilof M3 fell into the 13C mid-range of mainland M3s from that same laboratory that averaged 21.60 plusminus 0.70 yr (n = 83), so we can assume insignificant marine isotope contamination that might distort the dating by differential marine fractionization of the 13C and 14C isotopes of carbon.



  1. Palombo, M. R. The World of Elephants. Proc. 1st Internatl. Cong. 486–491 (Comune Di Roma, Consiglio Nazionale della Ricerche, Rome, 2001)
  2. Sondaar, P. Y. in Major Patterns of Vertebrate Evolution (eds Hecht, M. K., Goody, P. C. & Hecht, M. B.) 671–707 (Plenum, New York, 1977)
  3. Agenbroad, L. D. Pygmy mammoths, Mammuthus exilis, from Channel Islands National Park, California (USA). Deinsea 9, 1–39 (2003)
  4. Vartanyan, S. L. , Garutt, V. E. & Sher, A. V. Holocene dwarf mammoths from Wrangel Island in the Siberian Arctic. Nature 362, 336–339 (1993) | Article |
  5. Manley, W. F. Postglacial Flooding of the Bering Land Bridge: A Geospatial Animation v1. INSTAAR 2002 left fencehttp://instaar.colorado.edu/QGISL/bering_land_bridgeright fence
  6. Guthrie, R. D. Rapid body size decline in Alaskan Pleistocene horses before extinction. Nature 426, 169–171 (2003) | Article |
  7. Falconer, D. S. An Introduction to Quantitative Genetics (Longman, Harlow, 1989)
  8. Frankham, R. Inbreeding and extinction: island populations. Conserv. Biol. 12, 665–675 (2000) | Article |
  9. Hopkins, D. M. & Einarsson, T. Pleistocene glaciation on St. George Is. Pribilof Islands. Science 152, 343–345 (1966) | ISI |
  10. Sulerzhitsky, L. D. & Romanenko, F. A. The 'twilight' of the mammoth fauna in the Asiatic Arctic. Ambio 28, 251–255 (1999) | ISI |
  11. Crossen, K. J. , Graham, R. W. , Veltre, D. W. & Yesner, D. Abstr. Ann. Geol. Soc. Am. Conf. NW Div. 424 (Geol. Soc. Am., Seattle, Washington, 2003)
  12. Scheffer, V. B. Rise and fall of a reindeer herd. Sci. Monthly 73, 356–362 (1951)
  13. Colinvaux, P. A. Historical ecology in Beringia: the south land bridge coast at St. Paul Island. Quat. Res. 16, 18–36 (1981) | Article | ISI |
  14. Guthrie, R. D. Frozen Fauna of the Mammoth Steppe (Univ. Chicago Press, Chicago, 1990)
  15. Ager, T. A. Late Quaternary vegetation and climate history of the Bering Land Bridge from St. Michael Island western Alaska. Quat. Res. 60, 17–31 (2003) | Article |
  16. Martin, P. S. in Quaternary Extinctions (eds Martin, P. S. & Klein, R. G.) 345–403 (Univ. Arizona Press, Tucson, 1984)
  17. Guthrie, R. D. in Quaternary Extinctions (eds Martin, P. S. & Klein, R. G.) 259–298 (Univ. Arizona Press, Tucson, 1984)
  18. Agenbroad, L. D. , Johnson, J. & Morris, D. Abstr. Am. Quat. Assoc. 17th AMQUA Biennial Meeting (Am. Quat. Assoc., Univ. Alaska, Anchorage, 2002)
  19. Fitzhugh, W. W. Global cultural change: new views of circumpolar lands and people. Anthro. News Natl Mus. Nat. Hist. 9, 1 (1997)
  20. Gerasimov, D. , Girya, E. , Pitulko, V. & Tikhonov, A. New Materials for the Interpretation of the Site Chertov Ovrag on the Wrangel Island [in Russian] 379–383 (Mat. 2nd Dikov's conf., Magadan, SVKNII, 2002)
  21. Fladmark, K. R. Routes: alternate migration corridors for early man in North America. Am. Antiq. 4, 55–69 (1977)
  22. Laws, R. M. , Parker, I. S. C. & Johnston, R. C. B. Elephants and their Habitats: the Ecology of Elephants in North Bunyoro, Uganda (Clarendon, Oxford, 1975)
  23. Lister, A. M. Epiphyseal fusion and postcranial age determination in woolly mammoth. Ann. Nat. Hist. Mus. Rotterdam 6, 79–87 (1999)

Supplementary Information

Supplementary information accompanies this paper.



I thank O. Geist and R. Carroll who collected these island fossils, and the curators of the UAM, USNM and ANMH. M. L. Guthrie edited the manuscript. A. Lister and A. J Stuart provided reviews, and they also corroborated the mammoth date using their NERC Grant. My dating projects have been funded by the NSF.


Competing interests statement

The author declares no competing financial interests.


These links to content published by NPG are automatically generated.


Early Americans: Land bridge of Duvanny Yar

Nature News and Views (18 Apr 1985)

Mammoths in miniature

Nature News and Views (25 Mar 1993)

See all 4 matches for News And Views