Explaining the Pleistocene Megafaunal Extinctions: Models, Chronologies, and Assumptions

Home , Australian megafauna, Genyornis, Ratite

Perspective

Explaining the Pleistocene megafaunal extinctions: Models, chronologies, and assumptions

Barry W. Brook* and David M. J. S. Bowman

Key Centre for Tropical Wildlife Management, Northern Territory University, Darwin 0909, Northern Territory, Australia

Understanding of the Pleistocene megafaunal extinctions has been advanced recently by the application of simulation models and new developments in geochronological dating. Together these have been used to posit a rapid demise of megafauna due to over-hunting by invading humans. However, we demonstrate that the results of these extinction models are highly sensitive to implicit assumptions concerning the degree of prey naivety to human hunters. In addition, we show that in Greater Australia, where the extinctions occurred well before the end of the last Ice Age (unlike the North American situation), estimates of the duration of coexistence between humans and megafauna remain imprecise. Contrary to recent claims, the existing data do not prove the ‘‘blitzkrieg’’ model of overkill.

he cause of the extinction of giant birds, extra-Australian counterparts) apparently due solely to over-hunting by humans (16, Treptiles, and mammals in the late Pleis- occurred independent of climate change. 17). However, in both these models the tocene is, for palaeobiology, what Fermat’s For this reason, the situation in Greater unstated assumption was complete and un- last theorem was for mathematics (1): a Australia and New Zealand are seen as changing prey naivety. Recent field research long-standing scientific puzzle that has important ‘‘independent tests’’ of the impact has shown that both contemporary marsu- captured the imagination of specialists of human colonization on megafauna (3). pial and eutherian prey populations can and nonspecialists alike (2). Debate about Although a residual extant megafauna develop vigilance toward novel predators the possible cause of the extinction has did survive the Pleistocene extinction event quite rapidly (21, 22), which raises the ques- continued for over 150 yr (3–5), stimulated (e.g., red kangaroo, bison, Asian elephant, tion: What would happen to the predictions by new fossil finds, dating techniques, and llama, etc.), the only continent on Earth of these models if this assumption was modes of analysis. The debate is not strictly where a diverse assemblage of megafauna relaxed? scientific, however, because it impacts on remains is Africa, which is also where mod- the broader understanding of the evolution- ern humans arose. The African ‘‘anomaly’’ Sensitivity to the Assumption of Prey Naivety. ary theater of early human cultures, the fate is typically explained by long-term coevolu- We ran a simplified (single-species) version of contemporary global biodiversity, and the tion of megafauna with humans such that of Alroy’s (16) overkill model to evaluate rights of indigenous hunters (4, 6, 7). In- the prey and predator are matched evenly, the sensitivity of its results to axiomatic prey deed, the demise of the megafauna is often thereby creating trophic equilibrium. By naivety. This was done by systematically packaged as an environmental morality tale contrast, the extra-African megafauna are varying the term z in the equation that (8, 9). No matter why the debate interests characterized as completely naive to the describes predation efficiency (see Support- people, there is no doubt that it is of great human predator and therefore vulnerable ing Text, which is published as supporting significance. to overkill and the disintegration of food information on the PNAS web site, www. A great variety of competing scenarios webs (8, 15). pnas.org, and Fig. 1 for details). To date, all have been proposed to explain the extinc- A recent series of articles, widely reported overkill models assume prey off-take is de- tion of the megafauna such as climate in the international media, have been her- scribed by a ‘‘type II functional response,’’ change, disease, altered habitat condition alded as a major breakthrough in this de- where z ϭ 1, implying complete naivety of (particularly due to the effects of landscape bate. First, a mathematical simulation prey (e.g., refs. 10 and 16); or else take the burning by humans), and the breakdown of model (16) was interpreted as showing the extremely unrealistic view that off-take is food webs (3, 10, 11), but the presently North American megafauna most likely suc- entirely independent of prey density (e.g., ascendant idea is the so-called overkill hy- cumbed to over-hunting after human colo- ref. 17). pothesis. This theory posits that extinctions nization at the end of the last Ice Age. This When a functional response consistent were exclusively a result of human hunting article built on an earlier modeling exercise with ‘‘predator-hardened’’ prey is used, the (12), with a popular variant being the most concerning New Zealand extinctions (17). probability of overkill rapidly decays (Fig. restrictive case (termed ‘‘blitzkrieg’’), where Second, a metaanalysis of megafaunal re- 1). Yet when we set z ϭ 1 (Alroy’s assump- the extinction phase occurred within a few mains (18) found a coincidence between tion), our prediction of the probability of thousand years of human colonization (9, megafaunal extinction and human coloniza- overkill (93%) actually exceeded Alroy’s Ϸ 13). In North America, human colonization tion of Greater Australia (19, 20) 40,000– results (81%, based on the 37 scenarios in coincided with the end of the last Ice Age, 50,000 calendar years before present (yr which he treated megafauna as an undiffer- stymieing the disentanglement of human B.P.). Given the significance and impact of entiated ecological unit, as we have done). from environmental causes (12, 14). None- these articles, we chose to examine their Our higher predicted probability of overkill theless, proponents of the overkill hypoth- assumptions by undertaking some addi- when z ϭ 1 reflects our conservative as- esis point out that the megafauna had sur- tional simulations and statistical analyses sumption that human population dynamics vived previous glacial cycles where there was and reflect on possible future avenues that were independent of prey density, in con- no human predation. In Greater Australia may advance this complicated problem. trast to Alroy’s coupling of these terms. (Australia and the islands of Tasmania and New Guinea) and New Zealand, by con- Modeling Overkill trast, extinction of the megafauna (which Recent computer modeling has been used *To whom correspondence should be addressed. E-mail: were generally more diminutive than their to ‘‘prove’’ that megafaunal extinction was [email protected].

14624–14627 ͉ PNAS ͉ November 12, 2002 ͉ vol. 99 ͉ no. 23 www.pnas.org͞cgi͞doi͞10.1073͞pnas.232126899 Downloaded by guest on October 2, 2021 critical importance of considering whether rect consequence of human activities such as all of the extinct megafauna were naive and habitat modifications caused by landscape incapable of adapting to a new predator. burning. For instance, the destruction of woody vegetation by burning has been pos- ‘‘Calibration’’ of Model Assumptions. Con- tulated to explain the extinction of the Pleis- temporary studies provide the only realistic tocene Australian giant flightless bird, opportunity to test assumptions about how Genyornis newtoni, and all other megafauna megafauna adapt to human predation (10). Ϸ48,000 yr B.P. (11). This theory is consis- For instance, Berger et al. (21) studied how tent with the absence of evidence of human the relationship between moose (megafau- predation of the Australian megafauna nal prey) and wolves (predator) changed in such as kill sites or specialized tools (ref. 7 response to the time period over which they and S. Wroe, J. Field, R. Fullagar, and L. S. had coexisted. Compared with populations Jermiin, unpublished data) with the carbon- in which wolves and moose have always isotope signature of Genyornis fossil egg coexisted, moose exposed to reintroduced shells that indicates that it ate woody (C3 wolves showed an initially elevated level of photosynthetic pathway) plant species, in Fig. 1. Declining probability of simulated overkill predation and lack of response to threaten- contrast to fossil egg shells of the surviving as the cause of the megafaunal extinctions in relation ing stimuli but also rapidly lost this naive ratite, the Emu (Dromaius noveahol- to the departure from maximal predation efﬁciency vulnerability (although the z parameter can- ϭ landiae), which also eats tropical grass spe- (when z 1 in the equation of functional response, not be calculated from their data). Berger et cies (C4 photosynthetic pathway), and the where predation rate ϭ FP z͞[G ϩ P z]; see ref. 47). al. (21) concluded that the megafauna that apparent destruction of tropical rainforest The parameter z can be interpreted as a proxy survived the end-Pleistocene extinctions in Ϸ of relative megafauna naivety and has been ﬁxed at in the Australian humid tropics 47,000 yr 1 in all previous megafaunal extinction models, im- North America may have been better able to B.P. (30). plying complete prey naivety. adapt to novel predation than the species We conclude that models alone cannot that went extinct. resolve the question of whether the In northern Australia, experience gained megafauna were hunted to extinction be- Alroy (16) conceded that his model failed from management programs designed to cause of the critical importance of numer- to explain the differential survival of ‘‘geo- control feral animals such as the introduced ous assumptions, many of which may never graphically overlapping, morphologically Asian water buffalo has the potential to be amenable to validation. However, the similar and phylogenetically related extinct provide invaluable insights into the reaction demonstration of a long period of coexist- and extant species’’ such as the North Amer- of megaherbivores to intense human hunt- ence would indicate some stabilization of ican extinct and extant bison. He speculated ing pressure. Ridpath and Waithman (23) the ecological relationship between that this may be related to the secondary and Bayliss and Yeomans (24) demon- megafauna and human hunters, whereas a importance of ecological factors that were strated that the cost of exterminating buf- short period of overlap could imply, inter not included in his model. Yet an alternative falo rises in a negative exponential fashion alia, a blitzkrieg leading to a complete and equally valid explanation of this failure as populations decline to low densities. This breakdown of the predator–prey system, or to discriminate extinct and extant species is due to both a stimulation of the repro- some other rapid anthropogenic impact may be because of his invariant assumption ductive and natural survival rates to ap- such as massive landscape burning. An ac- of a type II functional response. These proach maximal levels (24, 25) and reduc- curate determination of the degree of over- anomalies, and the acknowledgment of tions in search efficiency. The latter lap between the extinct megafauna and hu- ‘‘gaps’’ in the model’s realism, underline the response reflects, in part, changed behavior man colonization is therefore clearly of of the animals such as increased wariness of great importance. This leads us next to

humans (23) and a shift from diurnal to consider recent advances in this field. PERSPECTIVE nocturnal grazing (26). An objection to inferences made from Dating the Greater Australian Record of such studies is that extant eutherian Humans and Megafauna megafauna such as buffalo and moose might No metaanalyses have tackled explicitly the remain genetically ‘‘hard-wired’’ to be wary question of the degree of overlap between of predators with which they have coevolved megafauna and humans. This issue remains (e.g., humans and wolves, respectively). highly contentious, and resolution is frus- However, some elegant recent studies have trated by the scarcity of suitable fossil and demonstrated that prey from a diverse range archaeological sites and the difficulty in of phylogenies can rapidly develop vigilance accurately dating them (18–20, 31, 32). to completely novel predators (see refs. 22, However, a number of recent attempts to 27, and 28), substantially weakening the resolve the individual issues of the earliest force of this criticism. colonization of Greater Australia by humans (19, 20) and the last occurrences of the The Limits of Modeling. The uncertainty of Greater Australian megafauna (18) repre- human–megafauna interactions is a great sent important steps in tidying up this messy impediment to developing realistic overkill field of inquiry. Here we consider the sta- Fig. 2. The sensitivity of the lower 95% conﬁdence models. For instance, if hunting of bound for the timing of the Australian megafaunal tistical robustness of the data for these megafauna was not primarily for subsistence datings (expressed in calendar years after extinction to the arbitrary truncation of the maxi- but for motives such as prestige or sexual mum age of articulated megafaunal skeletons in- appropriate calibrations). cluded in the analyses, assuming a uniform distribu- selection (e.g., ref. 29), then hunting pres- tion through time. The two youngest sites had an sure would be far higher than has been When Did the Greater Australian Megafauna identical age, necessitating a minimum inclusion of assumed in many models, and the probabil- Become Extinct? Recently, Roberts et al. (18) the three youngest sites. Roberts et al. (18) used a ity of overkill would be increased greatly undertook a metaanalysis of the existing truncation age of 55,000 yr B.P. (n ϭ 7 sites). (10). Conversely, extinction may be an indi- ‘‘reliable’’ data from 19 sites in Greater

Brook and Bowman PNAS ͉ November 12, 2002 ͉ vol. 99 ͉ no. 23 ͉ 14625 Downloaded by guest on October 2, 2021 Australia and concluded that the shown potential; see ref. 34). Is this a rea- skeleton from a sample of six sites ranges megafauna went extinct sometime between sonable stricture? from 0 to 2 to 43%. This uncertainty makes 51,200 and 39,800 yr B.P., with a most likely it presently impossible to discriminate con- Are Disarticulated Remains Necessarily Sus- date of 46,400 yr B.P. For their numerical fidently between alternative hypotheses. pect? analysis, they rejected all existing radiocar- A serious point of contention with the A possible test of the assumption of Rob- bon dates as too problematic, accepted only Roberts et al. (18) article was their decision erts et al. (18) could be to determine those skeletal remains that showed no evi- to treat all sites with disarticulated remains whether the likelihood of articulated fossils dence of anatomical disarticulation, and ex- as suspect, thereby excluding them from of extant species (e.g., kangaroos, emus, their analysis (34). What are the conse- cluded any dates older than 55,000 yr B.P., etc.) being preserved remained relatively quences of rejecting all the disarticulated leaving a sample size of seven sites with uniform through time for fossil deposits in a specimens? articulated remains. They believed these variety of environmental settings (e.g., hu- Our statistical analysis of the total data seven sites provided a statistically robust mid and arid zones). If there is no substan- set compiled by Roberts et al. (18) shows sample of the terminal extinction event. tial change in the ratio after the arrival of clearly that the sample of most recent ages A significant feature of the Roberts et al. humans, then this would support the con- from sites that contain disarticulated re- (18) article is that they provided a 95% tentious decision (32) to exclude disarticu- mains (n ϭ 8) are substantially younger than confidence interval around their estimate of lated remains from sites in which only the those that contain articulated skeletons (n ϭ the extinction event based on a compilation surrounding sediments have been dated 19); the median optically stimulated lumi- of dates rather than just using the measure- (e.g., Cuddie Springs) and remove the ques- nescence age (disarticulated) ϭ 37,500 yr ment error associated with the laboratory tion mark over the timing of the megafaunal B.P. versus articulated ϭ 63,000 yr B.P. A estimate of a single youngest date, as has extinction event. If, however, a substantial randomization test (35) shows this differ- become customary in Australia (31). How- change in the ratio did occur subsequent to ence to be highly significant (P Ͻ 0.001). ever, the assumptions driving their data human colonization, then the precise dating Further, the ratio of articulated to disartic- pruning and the small sample size that re- of the megafaunal extinction becomes ulated bones older than 46,000 yr B.P. is sults have been criticized (refs. 32 and 33 highly problematic and intractable until di- Ϸ10:1, whereas none of the younger sites and S. Wroe, J. Field, R. Fullagar, and L. S. rect dating of fossil remains becomes prac- contain articulated material. It should be Jermiin, unpublished data), prompting the tical and routine. question: How sensitive are their conclu- noted, however, that the true pre-46,000-yr- sions to these assumptions? old ratio may be close to or much greater When Did Humans First Arrive in Greater than unity because of the decision by Rob- Australia? The timing of human coloniza- What Is the Consequence of Data Pruning? erts et al. (18) to focus preferentially on tion is disputed vigorously (31). O’Connell We used a simulation method akin to boot- articulated skeletons. and Allen (19) argued that human coloni- strapping (see Supporting Text) to determine There are at least two explanations for zation happened no earlier than 43,000 yr Ͻ the lower age bound that resulted in a 5% the observed absence of articulated remains B.P. based on established radiocarbon- probability of obtaining a sample of size n more recent than 46,000 yr old. First, it dating procedures (conventional and accel- containing no date younger than 46,000 yr could be, as postulated by Roberts et al. (18), erator mass spectrometry methods with no old. An important feature of this approach that the supply of articulated megafauna advanced sample preparation) of archaeo- is that it seeks to identify the probable age skeletons ceased sometime after 46,000 yr logical material, a figure that also happens of the oldest remains by analyzing the age B.P. because the animals had become ex- to be a limit of the technique (39). However, distribution of all existing fossil finds, tinct. Alternatively, the megafauna may based on new advances in the removal of thereby obviating the futile quest to locate have actually persisted long after this date, contaminants of radiocarbon-dated sam- and date the very oldest site. However, these but the probability of a skeleton being pre- ples, Gillespie (20) disputes this conclusion predictions cannot overcome the inherent served in an articulated state substantially and suggests that colonization occurred laboratory measurement error associated declined over time, possibly because hu- somewhere between 5,000 and 10,000 yr with geochronology. mans were dismembering megafauna (e.g., earlier than the previously accepted radio- Our analysis shows that the results of by butchering or scavenging); however, an- carbon-date event horizon. Other new geo- Roberts et al. (18) are extremely robust thropological observations of modern indig- chronological techniques such as lumines- irrespective of their data pruning (Fig. 2). enous hunters report a positive relationship cence dating of quartz and uranium͞ Only when the three oldest sites are in- between prey body size and the likelihood thorium-series dating have apparently cluded in the sample does the 95% confi- that at least some skeletal elements will be pushed the date of human colonization to dence interval differ substantially from their left in an articulated state (36, 37). Equally, somewhere between 53,000 and 61,000 yr estimate. Indeed, our results show that the humans may have coincided with or caused B.P. (40, 41), although the veracity of these estimates derived from the existing data environmental changes that reduced the dates is still debated (e.g., refs. 19, 42, already fall within the bounds of the inher- likelihood of at least partially articulated and 43). ent laboratory measurement errors, and skeletons being preserved. This could be The uncertainty surrounding the validity thus additional dating of new fossil deposits due to increasing aridity (ref. 35 and S. of each of these advanced techniques makes will not be able to reduce this uncertainty Wroe, J. Field, R. Fullagar, and L. S. Jer- it impossible to decide which samples should further. We conclude that it is reasonable to miin, unpublished data) or some other fac- be retained or discarded. Indeed our re- state that the megafauna became extinct tor that changed geomorphological pro- Ϯ sampling analyses bracket the date of hu- sometime before 38,000–41,000 yr B.P. ( cesses such as accelerated soil erosion after man colonization of Greater Australia the estimated laboratory measurement er- sustained landscape burning. at between 44,200 and 71,500 yr B.P. (see ror). However, this conclusion depends crit- Which of these alternative explanations is Supporting Text). Obviously, such ambiguity ically on the assumption that articulated more plausible? Bootstrap resampling sim- has serious consequences for determining megafaunal skeletons provide the only prac- ulations (38) show that depending on as- the degree of overlap between humans and tical way of reliably dating extinctions, be- sumptions one makes about how the ratio of megafauna. cause reliable and routine direct dating of articulated to disarticulated remains have bones is currently outside the reach of the changed over time (e.g., constant at 10:1 or How Long Did Humans and Megafauna Co- available suite of geochronological tech- changing to either 1:1 or 1:10), the proba- exist in Greater Australia? Inherent measure- niques (although dating of tooth enamel has bility of failing to uncover an articulated ment errors associated with all geochrono-

14626 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.232126899 Brook and Bowman Downloaded by guest on October 2, 2021 logical techniques mean that the uncertainty analyzed after improved sample preparation of the superbly dated North American surrounding each individual date makes it tion, as has been advocated by Jones (44) record (e.g., refs. 4 and 14). Stable isotope presently impossible to determine whether and Roberts et al. (45). Yet even acid͞base͞ analyses of well dated fossil remains such as the overlap was less than the ‘‘radiometri- dichromate oxidation with stepped combus- those used by Miller et al. (11) should yield cally instantaneous interval’’ (for which tion radiocarbon dating has an upper tem- important insights into the ecology and eco- their mean dated ages must be separated by poral limit (of Ϸ50,000–55,000 yr B.P.; see physiology of Pleistocene fauna, thereby at least three times the sum of their standard refs. 20 and 34), thus constraining the utility permitting the testing and development of deviations; see ref. 31) regardless of the of the method in this situation. Finally, if the more refined hypotheses about the extinc- number of reliably dated samples. This pre- oldest archaeological dates based on the tion event. cludes direct dating of ‘‘geologically instan- more controversial nonradiocarbon meth- Quantitative modeling can investigate taneous’’ events such as a human-induced odologies are indeed true, we predict a 5% various scenarios by integrating existing rapid extinction, such as is required to dem- probability that coexistence was as long as knowledge in a logical, structured, and onstrate a blitzkrieg or massive ecological 25,500 yr (see Supporting Text for details). In transparent way, but it must be accepted disruption associated with anthropogenic sum, until the reliability of methods used to that the output of these models is a product landscape burning. However, taking the date the oldest archaeologist sites can be of explicit and implicit assumptions. Fur- extremes of the archaeological estimates agreed on, it is difficult to conceive how this ther, some explanations for extinction such for human colonization, it is at least pos- window of overlap can be determined more as habitat modification, cultural motives for sible to discriminate between two mutually precisely. hunting large game other than for subsis- exclusive alternatives: either there was no tence, climate change, and their interactions Conclusions overlap, or there was a long period of with hunting pressure are probably too com- coexistence. Inferring robustly the cause of the extinction plex to falsify with simulation modeling. If we take one extreme assumption, that of the Pleistocene megafauna is a remark- These caveats should be borne in mind when all the dates derived from the advanced ably complicated problem that is very sen- making claims about the causes of past geochronological methods (including acid͞ sitive to assumptions concerning the analy- ͞ extinctions, particularly given the way such base dichromate oxidation with stepped sis and interpretation of existing data. debates can influence current environmen- combustion radiocarbon dates) are invalid, Although great progress has been made, it is tal management policy (see refs. 4 and 46). then simulations show that there is a less premature to suggest (12, 16) that the prob- than 5% probability of any overlap between lem has been cracked. On the balance of This article builds on the constructive discussions humans and megafauna. Nevertheless, to current evidence, it is most unlikely that raised during a specialist workshop D.M.J.S.B. take this view is nonsensical. On the one there was no overlap between humans and attended on Australian megafaunal extinctions hand there has been no dispute of the use of megafauna, because for this to be true, all convened by Michael Smith of the National Mu- advanced geochronological techniques for advanced geochronological techniques must seum of Australia in September 2001. We thank dating the megafaunal extinction, but on the be rejected as erroneous. In addition, field Richard Roberts and Richard Gillespie for gen- other hand the application of these same and͞or laboratory tests are required to val- erously sharing their knowledge and unpublished techniques for archaeological materials is idate the assumption that the inferred age of background materials. The article was improved considered unreliable. If we more rationally all disarticulated fossils are unreliable. Al- greatly by the comments of Peter Bayliss, Graeme Brook, David Choquenot, Judith Field, Richard accept the radiocarbon dating based on though an improved chronology of the Gillespie, Geoff Hope, Jim O’Connell, Richard advanced sample preparation as valid, then Greater Australian archaeological record Roberts, Peter Whitehead, and Stephen Wroe. coexistence is absolutely certain, and the will eliminate some possible explanations, This work was undertaken by the Key Centre for period of overlap falls between 2,000 and dating alone cannot solve the problem of the Tropical Wildlife Management, Northern Terri- 10,000 yr. However, this time interval may cause of the extinctions, as evidenced by the tory University, under funding from the Austra- change as more archaeological sites are re- ongoing debate concerning the interpreta- lian Research Council. PERSPECTIVE

1. Singh, S. (1998) Fermat’s Last Theorem: The Story of a Riddle 18. Roberts, R. G., Flannery, T. F., Ayliffe, L. K., Yoshida, H., 34. Turney, C. S. M., Bird, M. I., Fifield, L. K., Roberts, R. G., That Confounded the World’s Greatest Minds for 358 Years Olley, J. M., Prideaux, G. J., Laslett, G. M., Baynes, A., Smith, Smith, M., Dortch, C. E., Grun, R., Lawson, E., Ayliffe, L. K., (Fourth Estate, London). M. A., Jones, R. & Smith, B. L. (2001) Science 292, 1888– Miller, G. H., Dortch, J. & Cresswell, R. G. (2001) Quaternary 2. Diamond, J. M. (2001) Nature 411, 755–757. 1892. Res. 55, 3–13. 3. Martin, P. S. & Klein, R. G. (1984) Quaternary Extinctions: A 19. O’Connell, J. F. & Allen, J. (1998) Evol. Anthropol. 6, 35. Manly, B. F. J. (1991) Randomization and Monte Carlo Prehistoric Revolution (Univ. of Arizona Press, Tucson). 132–146. Methods in Biology (Chapman and Hall, London). 4. Grayson, D. K. (2002) Bull. Fla. Mus. Nat. Hist., in press. 20. Gillespie, R. (2002) Radiocarbon, in press. 36. Bunn, H. T., Bartram, L. & Kroll, E. (1988) J. Anthropol. 5. Martin, P. S. & Steadman, D. W. (1999) in Extinctions in Near 21. Berger, J., Swenson, J. E. & Persson, I. (2001) Science 291, Archaeol. 7, 412–457. Time, ed. MacPhee, R. D. E. (Kluwer Academic, New York), 1036–1039. 37. O’Connell, J. F., Hawkes, K. & Blurton-Jones, N. G. (1992) pp. 17–52. 22. McLean, I. G., Lundie-Jenkins, G. & Jarman, P. J. (1996) J. Archaeol. Sci. 19, 319–345. 6. Bowman, D. (2001) J. Biogeogr. 28, 549–564. Biol. Conserv. 75, 51–62. 38. Efron, B. & Tibshirani, R. J. (1993) An Introduction to the 7. Wroe, S., Field, J. & Fullagar, R. (2002) Nat. Aust. 27, 54–61. 23. Ridpath, M. G. & Waithman, J. (1988) Wildl. Soc. Bull. 16, Bootstrap (Chapman & Hall, New York). 8. Flannery, T. F. (2001) The Eternal Frontier: An Ecological 385–390. 39. Chappell, J., Head, J. & Magee, J. (1996) Antiquity 70, 16, History of North America and Its People (Atlantic Monthly, 24. Bayliss, P. & Yeomans, K. M. (1989) Aust. Wildl. Res. 543–552. 651–676. 40. Roberts, R. G., Jones, R., Spooner, N. A., Head, M. J., New York). 25. Boulton, W. J. & Freeland, W. J. (1991) Wildl. Res. 18, 63–73. Murray, A. S. & Smith, M. A. (1994) Quaternary Sci. Rev. 13, 9. Flannery, T. F. (1994) The Future Eaters: An Ecological 26. Freeland, W. J. & Bolton, W. J. (1990) Aust. Wildl. Res. 17, 575–583. History of the Australasian Lands and People (Reed Books, 411–420. 41. Thorne, A., Grun, R., Mortimer, G., Spooner, N. A., Simp- Sydney). 27. Griffin, A. S., Blumstein, D. T. & Evans, C. (2000) Conserv. son, J. J., McCulloch, M., Taylor, L. & Curnoe, D. (1999) J. 10. Choquenot, D. & Bowman, D. M. J. S. (1998) Global Ecol. Biol. 14, 1317–1326. Hum. Evol. 36, 591–612. 7, Biogeogr. Lett. 167–180. 28. Blumstein, D. T., Daniel, J. C., Griffin, A. S. & Evans, C. S. 42. Allen, J. & Holdaway, S. (1995) Antiquity 69, 101–112. 11. Miller, G. H., Magee, J. W., Johnson, B. J., Fogel, M. L., (2000) Behav. Ecol. 11, 528–535. 43. Bowler, J. M. & Magee, J. W. (2000) J. Hum. Evol. 38, Spooner, N. A., McCulloch, M. T. & Ayliffe, L. K. (1999) 29. Hawkes, K., O’Connell, J. F. & Rogers, L. (1997) Trends Ecol. 719–726. Science 283, 205–208. Evol. 12, 29–32. 44. Jones, R. (1982) in Archaeometry: An Australasian Perspective, 12. Flannery, T. F. (1999) Science 283, 182–183. 30. Turney, C. S. M., Kershaw, A. P., Moss, P., Bird, M. I., Fifield, eds. Ambrose, W. & Duerden, P. (Dept. of Prehist., Res. Sch. 13. Martin, P. S. (1973) Science 179, 969–974. L. K., Cresswell, R. G., Santos, G. M., Di Tada, M. L., Pac. Stud., Australian Natl. Univ., Occas. Pap. Prehist. 12, 14. Beck, M. W. (1996) Paleobiology 22, 91–103. Hausladen, P. A. & Zhou, Y. (2001) J. Quaternary Sci. 16, Canberra). 15. Martin, P. S. (1967) in Pleistocene Extinctions: The Search for 767–771. 45. Roberts, R. G., Jones, R. & Smith, M. A. (1994) Antiquity 68, a Cause, eds. Martin, P. S. & Wright, H. E. (Yale Univ. Press, 31. Webb, R. E. (1998) Radiocarbon 40, 749–758. 611–616. New Haven, CT), pp. 75–120. 32. Field, J., Fullager, R., Roberts, R. G., Yoshida, H., Flannery, 46. Jackson, J. B. C., Kirby, M. X., Berger, W. H., Bjorndal, K. A., 16. Alroy, J. (2001) Science 292, 1893–1896. T. F., Ayliffe, L. K., Olley, J., Prideaux, G. J., Laslett, G., Botsford, L. W., Bourque, B. J., Bradbury, R. H., Cooke, R., 17. Holdaway, R. N. & Jacomb, C. (2000) Science 287, 2250– Baynes, A., et al. (2001) Science 294, 7. Erlandson, J., Estes, J. A., et al. (2001) Science 293, 629–638. 2254. 33. Wroe, S. & Field, J. (2001) Australasian Sci. 22, 21–25. 47. Real, L. A. (1977) Am. Nat. 111, 289–300.

Brook and Bowman PNAS ͉ November 12, 2002 ͉ vol. 99 ͉ no. 23 ͉ 14627 Downloaded by guest on October 2, 2021