R EPORTS on the bird’s extinction within each region. By dating the extinction event in several re- Pleistocene Extinction of gions spanning a range of climates, continen- tal extinction may be evaluated. Genyornis newtoni: Human Most birds incorporate about 3% complex organic compounds in their eggshells as an Impact on Australian intracrystalline organic web that is highly resistant to diffusional loss. Epimerization of the common protein amino acid isoleucine, Megafauna preserved in the proteinaceous residues, fol- Gifford H. Miller,* John W. Magee, Beverly J. Johnson,† lows first-order, reversible kinetics (9, 10). Expressed as the proportion of the non-pro- Marilyn L. Fogel, Nigel A. Spooner, Malcolm T. McCulloch, tein D-alloisoleucine to its protein source, ‡ Linda K. Ayliffe L-isoleucine, D/L increases from essentially zero in modern eggshell to an equilibrium More than 85 percent of Australian terrestrial genera with a body mass ex- ratio of 1.30 (11). The temperature depen- ceeding 44 kilograms became extinct in the Late Pleistocene. Although most dence of isoleucine epimerization in Drom- were marsupials, the list includes the large, flightless mihirung Genyornis new- aius eggshell is well known (12). Eggshells toni. More than 700 dates on Genyornis eggshells from three different climate of both Genyornis and Dromaius commonly regions document the continuous presence of Genyornis from more than occur together in all regions (13), suggesting 100,000 years ago until their sudden disappearance 50,000 years ago, about the that they coexisted and nested in close prox- same time that humans arrived in Australia. Simultaneous extinction of Ge- imity, contrary to a recent assertion (14). nyornis at all sites during an interval of modest climate change implies that We measured D/L ratios in eggshells of human impact, not climate, was responsible. both taxa from three regions (15): (i) an extensive series (1200 analyses) from dune Australia suffered a major loss of its large- but more recently has included indirect con- and beach deposits within a 50-km radius of and medium-sized land mammals in the Late sequences of human activity, particularly Madigan Gulf (MG) in Lake Eyre, the termi- Pleistocene. All marsupials exceeding 100 kg ecosystem change resulting from burning nal playa of the Lake Eyre Basin; and modest (19 species) and 22 of 38 species between 10 practices (6–8). Resolving the debate requires collections (ii) from a 50-km radius centered on March 4, 2009 and 100 kg became extinct (1), along with secure dates on the extinction events, on the on the north end of Lake Frome (LF) and (iii) three large reptiles and the ostrich-sized Ge- arrival of humans in Australia, and on major from a 150-km radius centered on the conflu- nyornis newtoni (2); two other large flightless climate and environmental changes. Here we ence of the Darling and Murray rivers (DM) birds, the emu (Dromaius novaehollandiae) present new dates that constrain the ages of (Fig. 1). The MG region is currently arid and the cassowary (Casuarius casuarius), these key events based on amino acid race- [21°C, ϳ150 mm precipitation (16)]; LF is survived. Dwarfing and range restriction oc- mization, accelerator mass spectrometry semiarid (19°C, ϳ400 mm precipitation) and curred in many other species (1). Collective- (AMS), 14C and thermal ionization mass has denser plant cover and thick riparian ly, the lost species are often referred to as the spectometry (TIMS) U-series analyses on vegetation; and the DM region is semiarid Australian megafauna, although most are of eggshells, and luminescence dates on associ- (17°C, ϳ400 mm precipitation) with more www.sciencemag.org modest body mass. ated sediment. reliable stream flow. The Darling River For more than a century the cause of this The abundance of fossil eggshell reflects, drains the summer-dominant precipitation re- exceptional extinction has been debated with- in order of importance, its preservation po- gime of the north, whereas the Murray drains out a clear consensus, largely because of the tential, subsequent exposure potential, and the winter-westerly–dominated precipitation difficulty in dating faunal remains close to the number of nesting birds. To be preserved, regime of southern Australia. Consequently, the limit of radiocarbon dating (3). The cause mobile sediment, commonly eolian sand, the DM sites have one of the most depend-

of megafauna extinction initially focused on must be available to bury the eggshell quick- able moisture supplies in the interior, and Downloaded from climate change (4) or human predation (5), ly. Eggshell is most frequently collected from should have been a refugium for fauna during natural exposures or deflation hollows. With- periods of climate stress. In all three regions, out recent erosion, eggshell is rarely encoun- precipitation can occur in any season, but G. H. Miller and B. J. Johnson, Center for Geochronical Research, Institute of Arctic and Alpine Research tered because it survives only briefly at the summer precipitation dominates in the MG (INSTAAR), and Department of Geological Sciences, surface. Bird populations are controlled pri- region. Before European settlement, it sup- University of Colorado, Boulder, CO 80309–0450, marily by climate, nutrient availability, and ported mixed C3 and C4 vegetation. LF and USA. J. W. Magee, Department of Geology, Australian predator pressure (including humans). DM support primarily C plants. National University, Canberra ACT 0200, Australia. 3 M. L. Fogel, Geophysical Laboratory, Carnegie Institu- Eggshells of large, flightless birds offer In the MG region, the D/L record is cali- tion of Washington, 5251 Broad Branch Road, NW, distinct advantages over bone for dating pur- brated by 90 paired AAR and calibrated AMS Washington, DC 20015, USA. N. A. Spooner, M. T. poses. They are well preserved and surpris- 14C dates (17)onDromaius eggshells, 8 lu- McCulloch, L. K. Ayliffe, Research School of Earth ingly common across Australia, their dense minescence dates between 40 and 120 thou- Sciences, Australian National University, Canberra, calcite matrix is resistant to diagenesis, and, sand years ago (ka) on eolian sand from ACT 0200, Australia. from a single fragment, 14C activity, amino which eggshell has been collected, and 5 *To whom correspondence should be addressed. E- acid racemization (AAR), and stable isotope TIMS U-series dates on Genyornis eggshell. mail: [email protected] †Present address: School of Oceanography, Box ratios can be measured. By dating eggshells An age model developed from this calibration 357940, University of Washington, Seattle, WA of Genyornis and Dromaius collected from set allows calendar ages to be assigned to 98195–7940, USA. deposits formed during wetter and drier in- each D/L ratio for all samples Ͻ130,000 years ‡Present address: Laboratoire des Sciences du Clima- tervals, environmental change may be placed old (Fig. 2). tet et de l’Environnement (LSCE), Domaine du CRNS– Bt 12, Ave. de la Terrasse, 91198 Gif-sur-Yvette, in a secure time frame, and the youngest Frequency histograms for the two taxa Cedex, France. Genyornis eggshells can hence provide a date derived from the age model document the

www.sciencemag.org SCIENCE VOL 283 8 JANUARY 1999 205 R EPORTS time dependence of the collection series over dated Genyornis eggshells directly by TIMS age of 37,600 14C yr B.P., and three eggshells the past 130 thousand years (ky) (Fig. 3); the U-series (18), and dated sediment enclosing from within the extinction window range be- frequency distribution of the 90 14C-dated the youngest Genyornis eggshell by lumines- tween 38,100 and Ͼ45,100 14C yr B.P. Two Dromaius eggshells (Fig. 3) mirrors that of cence (19). At MG, Genyornis became ex- Genyornis eggshells that slightly predate the the larger data set of model-derived ages. MG tinct between D/L ratios of 0.46 and 0.50. extinction, based on their D/L ratios, yielded was continually occupied by breeding popu- AMS 14C dates on 11 Dromaius eggshells ages of 42,600 and Ͼ44,400 14C yr B.P., and lations of Dromaius for at least the past 130 within that range are all Ͼ38,700 14C years five fragments from a single stratigraphic level, ky, even through the stage 2 arid phase. before the present (yr B.P.), essentially at the all with D/L ratios of 0.34 and therefore slightly Intervals of greatest Dromaius eggshell abun- limit of the method. Five luminescence dates older than the extinction, have an average dance reflect peak sediment mobility (preser- from the corresponding eolian unit are be- TIMS U-series age of 53 Ϯ 3ka(22). vation) during early stage 5, early stage 3, tween 63 Ϯ 5 ka and 44 Ϯ 5ka(20). U-series No independent dates are available from and at about the stage 2/1 transition. Al- dates on terminal Genyornis eggshells from LF. The temperature averages 2°C lower though the distribution of Genyornis egg- two different sites in MG are 47 and 65 ka. there than at MG (21); consequently, the D/L shells is similar for the older part of the Radiocarbon ages of Genyornis eggshell that ratio expected of a correlative extinction record, eggshells abruptly disappear about received only an acid-leaching pretreatment event is 0.40 Ϯ 0.03, midway between those 50,000 years ago. The continuous presence of suggest that many of these eggshells are con- of extinction events at MG and LF. The D/L Dromaius eggshells confirms that the ab- taminated by younger carbon. For example, ratios of the youngest Genyornis at LF are sence of younger Genyornis eggshell is not three fragments from levels stratigraphically higher (0.46 Ϯ 0.02); hence, extinction may due to a lack of preservation or exposure below luminescence dates of 50 to 60 ka have occurred slightly earlier, although the potential, but must indicate the actual extinc- produced finite apparent radiocarbon ages be- small sample size may mean we failed to tion of the birds in this area. tween 36,700 and 39,500 14C yr B.P. Using a locate the youngest Genyornis there. Insufficient age calibration is available for more rigorous pretreatment (90% acid leach; Our dates bracket the timing of Genyornis the other two regions to plot frequency distri- drilling out all pores), we obtained radiocar- extinction. Luminescence and U-series dates butions in the time domain, but we can evaluate bon ages older than 42,000 14C yr B.P. for demonstrate that Genyornis was common as them in the D/L domain (Fig. 4). D/L is a two fragments of Genyornis that gave the recently as 60 to 50 ka. The radiocarbon time-dependent quantity, but the epimerization lowest D/L in the MG collections and that evidence from eggshells of both taxa span- rate exhibits a strong temperature dependency, previously gave finite radiocarbon ages ning the extinction event collectively demon- and Pleistocene temperatures in central Austra- younger than 40,000 14C yr B.P. with less strate that it occurred at or beyond the effec- lia were up to 9°C lower than in the Holocene rigorous pretreatment. We conclude that the tive limit of 14C dating of carbonates (at least on March 4, 2009 (12); consequently, time is not linearly ex- extinction event occurred at or beyond the 42,000 14C yr B.P., corresponding to a cal- pressed. Furthermore, because the LF and DM limits of 14C dating. endar age of Ͼ46 ka). Thus, the extinction regions are cooler than the MG region, the age In the DM region Genyornis became extinct for a specific D/L ratio is higher. Within these at a D/L ratio between 0.28 and 0.32, substan- constraints, D/L histograms for all three regions tially lower than at MG because average tem- exhibit similar trends. All three reveal near peratures at DM are 4°C less (21). One continuous occupation by Dromaius. The lack Dromaius eggshell slightly younger than the of old material at DM simply reflects the poor extinction event (D/L ϭ 0.27) has a radiocarbon exposure there of older sediment. In sharp con- www.sciencemag.org trast to the continuity of Dromaius histograms, Genyornis displays a distinct termination in all three areas. Because of the temperature differences, the age of the extinction event must be eval- uated at each site. To do this, we radiocarbon

dated fragments of both Genyornis and Downloaded from Dromaius that straddle the extinction event,

Fig. 2. Age model used to convert D/L ratios to calendar ages based on 90 paired and calibrated (17) AMS 14C dates and D/L ratios, 5 paired TIMS U-series dates and D/L ratios, and 8 lumi- nescence dates on sediment from which egg- shell D/L ratios were measured. The two initial Fig. 3. Age-frequency histograms for Dromaius linear segments are based on earlier assess- (above) and Genyornis (below) for the past 130 ments (12), whereas the older two linear seg- ky from the MG region, derived from D/L ratios ments are simple regressions fit through the on individual eggshell fragments collected from data. Other configurations are plausible, but widely scattered sites that have been converted yield similar conclusions. LGM deposits in MG to calendar ages from the age model shown in are thin, and the shallow burial, coupled with Fig. 2. Solid bars in the upper panel represent Fig. 1. Locality map showing the Lake Eyre the low glacial-age temperature, results in poor the AMS 14C-dated eggshells superimposed on Basin (hatched area), Madigan Gulf sites (MG), age resolution for the interval between 16 and the AAR-dated series. The sudden disappear- Lake Frome sites (LF), and the Darling-Murray 60 ka, but the similar patterns for 14C- and ance of Genyornis ϳ50 ka defines the extinc- sites (DM) at the confluence of the Darling (D) AAR-dated samples (Fig. 3) suggest that any tion event in this region. Marine isotope stage and Murray (M) rivers. biases in the histograms are minimal. boundaries are shown.

206 8 JANUARY 1999 VOL 283 SCIENCE www.sciencemag.org R EPORTS occurred 50 Ϯ 5 ka. The similar pattern and than at 140 ka, and probably less than the arid able dates are luminescence ages of quartz timing of Genyornis extinction across three dif- phase 30 to 16 ka. Eolian sands were depos- grains enclosing human artifacts. The oldest ferent regions, including the DM region that ited adjacent to the lake during the highstand widely accepted dates are 55 Ϯ 5ka(25, 26), would be expected to serve as a refugium dur- 60 ka and as it subsequently dried. Genyornis near the time of the extinction; older dates ing times of drought, and the continuous occur- eggshells occur throughout the eolian unit, from northwest Australia (27) have been rence of Dromaius through the extinction event but not in younger deposits, indicating that questioned (28, 29), and indirect evidence and on to the present at all three sites suggest extinction occurred at the end of a modest that has been interpreted as humanly induced, that extinction was taxonomically selective and drying event or during an immediately sub- such as changes in fire regime (30, 31), re- on a continental scale. sequent wetter interval. mains tenuous. To evaluate whether a change in climate The most complete record of moisture Humans are known to have had an impact may have led to the extinction, we examined change in the DM region comes from the Wil- on large birds elsewhere. The extinction of all paleoclimate reconstructions that extend landra Lakes (24). A long period of aridity species of moa (Dinornis spp.) from New through the extinction event in MG (past 150 terminated about 60 ka, with increased moisture Zealand within a few centuries of human ky) (20, 23) and DM (past 70 ky) (24), con- supporting abundant vegetation and filling the colonization is documented to have been a centrating on changes in available moisture, basin with permanent water. Genyornis and consequence of hunting and fires (32). Geny- the limiting variable for vegetation across Dromaius eggshells are present in deposits of ornis carried about twice the body mass of most of Australia. this age. Beginning at ϳ40 ka, the climate Dromaius and was presumably less agile. It is During arid times, when precipitation is re- began to become drier; Dromaius was present, plausible that because it was less mobile, or duced over the Lake Eyre Basin, playa condi- but no Genyornis have been reported in this or possibly because of a more vulnerable breed- tions dominate in MG, and groundwater salinity any younger units. After an interval of variable ing strategy, Genyornis was selectively pre- rises because of evaporative concentration. As conditions, final desiccation occurred shortly dated by early hunters. However, evidence of the watertable is lowered, salts that crystallize at after 20 ka, and Lake Mungo has remained dry direct predation on Genyornis by humans is the groundwater capillary fringe within the sed- since. In this region, Genyornis apparently be- limited to a single site (33), and kill sites for iment matrix disrupt the surface and promote came extinct before the onset of aridity. other elements of the megafauna are equally deflation, removing sediment and salts from the It thus seems that Genyornis was able to rare [for example, (1, 34–36)]. Alternatively, basin. While disruption and deflation keep pace survive through the range of natural environ- ecosystem disruption as a consequence of with watertable lowering, the playa floor is mental changes caused by Pleistocene cli- early human activity may have impacted the lowered, until the groundwater level reaches mate oscillations. At the time of Genyornis fauna differentially, possibly on the basis of equilibrium with the evaporative conditions, extinction, 50 Ϯ 5 ka, modest aridity (MG) to dietary preferences. on March 4, 2009 and a halite crust develops. Thus, the maximum slightly wetter-than-present conditions (DM) The gross feeding habits of Genyornis can depth of deflation is a secure proxy for the prevailed. Consequently, climate change as be reconstructed from the isotopic composition intensity of an arid phase. Arid-phase defla- an explanation for Genyornis extinction is of its eggshells, and its skeletal morphology. tion excavated the playa floor below its cur- unlikely. Genyornis is considered to be an The 13C/12C ratio of carbon in a bird’s food rent level (15 m below sea level) about 140 element of the Australian megafauna (1), and source is preserved in the calcite matrix of, and ka, 60 to 50 ka, and between about 30 and 16 its bones are associated with some of the organic compounds within, its eggshells. In ka. Of these, the most extensive excavation youngest stratigraphic levels containing modern eggshell of Dromaius, Struthio (os- occurred at 140 ka, when the playa was erod- megafaunal remains. Thus, we argue that ex- trich), and Coturnix (quail), the ␦13Cofthe ed to bedrock, as much as 19 m below sea tinction of the Australian megafauna in gen- organic fraction is enriched by 2 Ϯ 1 per mil www.sciencemag.org level. Lake Eyre last carried a permanent eral is likely to be coincident with Genyornis relative to the diet of the parent; the offset waterbody ϳ60 ka; slow drying followed, but extinction, and that this occurred ϳ50 ka. between the organic and inorganic fractions is the extent of playa deflation, and by direct Dating the arrival of the first humans to 9 Ϯ 1, 15 Ϯ 1, and 14 Ϯ 1 per mil, respectively inference the intensity of aridity, was less Australia has proven difficult. The most reli- (37, 38). The organic-inorganic offsets are in- distinguishable in Late Quaternary eggshell of Dromaius (10 Ϯ 2 per mil; n ϭ 45) and Stru- Ϯ ϭ

thio (14 1 per mil; n 49). The average Downloaded from isotopic offset between the two fractions for fossil Genyornis (14 Ϯ 2 per mil; n ϭ 74) is similar to that of Struthio and Coturnix.Ifwe assume that the same 2 per mil enrichment of the eggshell organic fraction over dietary ␦13C measured in our modern controls applies to fossil samples, it is possible to reconstruct di- etary ␦13C for the extinct Genyornis. Dietary ␦13C reconstructed from 65 Geny- ornis and 49 Dromaius eggshells that were laid between 60 and 45 ka in MG averages Ϫ22 and Ϫ18 per mil, respectively. Twenty of 49 Dromaius eggshells are between Ϫ13 and Ϫ15 per mil, but some individuals have values as light as Ϫ26 and as heavy as Ϫ9 per mil. Genyornis group more tightly (54 of 65 are Fig. 4. D/L-frequency histograms for Dromaius and Genyornis from all three regions. Dromaius D/L between Ϫ19 and Ϫ25 per mil; none are heavi- ratios have been converted to equivalent Genyornis D/L ratios by dividing by 1.18, their average Ϫ offset in paired collections of the two taxa, so that within each region the two taxa can be directly er than 14 per mil), suggesting that Genyornis compared. Some key radiocarbon dates are indicated at their associated D/L ratio, showing that ate primarily C3 browse, whereas Dromaius extinction occurred at or beyond the limit of 14C dating of eggshell. had wider dietary tolerances including a signif-

www.sciencemag.org SCIENCE VOL 283 8 JANUARY 1999 207 R EPORTS

icant component of C4 grasses. In contrast, 25 3. Direct dating of megafaunal bone by radiocarbon fossil samples these have been infilled with clay, a MG Dromaius eggshells from the last glacial remains hampered by the poor preservation of colla- carrier of thorium. We thus restrict our discussion to 230 232 Ͼ maximum (AMS 14C dates for each are be- gen, and most dates on faunal remains are close to samples with Th/ Th activity ratios 30. the radiocarbon limit. Although collagen constitutes 19. Supplementary material for 14C, TIMS U-series, and tween 28,000 and 14,000 yr B.P.) have a di- 25% by weight in modern bone, under alternately OSL data is available at www.sciencemag.org/ etary ␦13C clustered between Ϫ19 and Ϫ22 per wet and dry conditions in warm regions like Australia feature/data/985798.shl. mil, indicating a shift to regional dominance by it is rapidly denatured into soluble gelatin and 20. J. W. Magee, thesis, Australian National University, C plants and a corresponding adjustment in the leached from the bone (41). Baynes (42) reviewed 91 Canberra (1997). 3 published radiocarbon dates on megafaunal remains 21. J. P. McMahon, M. F. Hutchinson, H. A. Nix, K. D. Ord, birds’ diet. Genyornis from the DM region or sediments containing them; ages as young as 6000 Centre for Resource and Environmental Studies, Aus- ␦13 ϭ 14 apparently ate exclusively C3 plants ( C C yr B.P. have been accepted by some authors. tralian National University, Canberra (1995). Ϫ26 Ϯ 1 per mil; n ϭ 9). After applying rigorous criteria to assess the reliabil- 22. Although the 230Th/232Th activity ratios of these samples are below our preferred threshold, the re- The isotopic data indicate that Dromaius ity of a radiocarbon date and its association with faunal remains (43), he rejected most dates, in- producibility between many samples from a single had broader dietary tolerances than did Ge- cluding all those younger than 28,000 14C yr B.P. level with large differences in U concentrations sug- nyornis and was able to adapt its feeding The few remaining sites with apparent finite 14C gests that the ages are reliable. 23. J. W. Magee, J. M. Bowler, G. H. Miller, D. L. G. ages between 28,000 and 40,000 yr B.P. were not strategies to a wider variety of ecological Williams, Palaeogeogr. Palaeoclimatol. Palaeoecol. directly on megafaunal remains, and either the conditions. This ability may have contributed 113, 3 (1995). association of the megafauna with the dated level 24. J. M. Bowler, Archaeol. Oceania 33, 120 (1998). to its survival through the extinction event. remains debated or there are uncertainties in the The cranial morphology of the dromorni- 25. R. G. Roberts, R. Jones, M. A. Smith, Nature 345, 153 source of the actual carbon that was dated, leaving (1990). thids, of which Genyornis is a member, indi- the timing of extinction unresolved. 26. R. G. Roberts et al., Quat. Geochronol. 13, 575 cates that the beak was used primarily in a 4. R. Tate, Trans. Proc. Philos. Soc. S. Aust. 2, 13 (1879). (1994). shearing motion, a characteristic of browsing 5. R. Owen, Researches on the Fossil Remains of the 27. R. L. K. Fullagar, D. M. Price, L. M. Head, Antiquity 70, Extinct Mammals of Australia (Erxleben, London, 751 (1996). animals (39). Consequently, we conclude that 1877). 28. N. A. Spooner, ibid. 72, 173 (1998). Genyornis was primarily a browser and likely 6. D. Merrilees, J. R. Soc. West. Aust. 51, 1 (1968). 29. R. G. Roberts et al., Nature 393, 358 (1998). to be dependent on extensive shrubland. 7. R. Jones, Aust. Nat. Hist. 16, 224 (1969). 30. G. Singh, A. P. Kershaw, R. Clark, in Fire and Australian ࿜࿜࿜࿜ Biota, A. M. Gill, R. A. Groves, I. R. Noble, Eds. (Aus- Fire has exerted a major control on Aus- 8. , Nature 246, 278 (1973). 9. A. S. Brooks et al., Science 248, 60 (1990). tralia Academy of Science, Canberra, 1981), pp. 23– tralian vegetation throughout the Quaternary, 10. G. H. Miller, P. B. Beaumont, A. J. T. Jull, B. J. Johnson, 54. but the adaptations to fire are to ignition Philos. Trans. R. Soc. London Ser. B 337, 149 (1992). 31. A. P. Kershaw, G. M McKenzie, A. McMinn, Proc. during a restricted portion of the year and 11. Epimerization rates exhibit systematic offsets be- Ocean Drilling Prog. Sci. Results 133, 107 (1993). tween taxa as a result of differences in the amino 32. M. M. Trotter and B. McCulloch, in Quaternary Ex- only after sufficient fuel loading had oc- tinctions: A Prehistoric Revolution, P. S. Martin and

acid sequences in the protein each bird uses. D/L on March 4, 2009 curred. One possibility is that burning prac- ratios measured in Dromaius and Genyornis eggshells R. G. Klein, Eds. (Univ. of Arizona Press, Tucson, from the same stratigraphic horizon indicate that 1984), pp. 708–727. tices of the earliest human immigrants dif- 33. M. Smith, G. Miller, G. Van Tets, Rec. S. Aust. Mus. 27, fered enough from that of the natural fire isoleucine epimerizes 1.18 times faster in Dromaius; Dromaius D/L ratios have been adjusted by this 228 (1994). cycle to disrupt ecosystems across the semi- amount to allow plots comparing the two taxa (Figs. 34. D. R. Horton, Archaeol. Phys. Anthropol. Oceania 15, arid zone. The ecosystem may have been 3 and 4) on a common scale. 86 (1980). 35. T. F. Flannery, The Future Eaters (Reed Books, Mel- particularly susceptible as a result of the con- 12. G. H. Miller, J. W. Magee, A. J. T. Jull, Nature 385, 241 (1997). bourne, 1994). tinent’s geological quiescence; soils in this 13. Dromaius and Genyornis apparently nested in similar 36. J. L. Kohen, Aboriginal Environmental Impacts (Univ. low-relief landscape were depleted of most habitats; eggshells of both taxa are commonly found of New South Wales Press, Sydney, 1995). at the same stratigraphic level in many eolian depos- 37. B. J. Johnson, thesis, University of Colorado (1995). nutrients, resulting in a lack of ecosystem 38. ࿜࿜࿜࿜, M. L. Fogel, G. H. Miller, Geochim. Cosmo- its, although Genyornis is usually the dominant taxon. www.sciencemag.org resiliency. We postulate that human burning Both are occasionally found as crushed whole or chim. Acta 62, 2451 (1998). at times of the year and at frequencies to partial eggs, more often as concentrations of eggshell 39. P. Murray and D. Megirian, Rec. S. Austr. Mus.,in which the vegetation was not pre-adapted fragments. Eggshells of the two genera can be readily press. distinguished by characteristic macro- and micro- 40. A. P. Kershaw, Nature 322, 47 (1986). resulted in a dramatic decrease in tree and structures (44). 41. T. W. J. Stafford, P. E. Hare, L. Currie, A. J. T. Jull, D. shrub vegetation across the continental inte- 14. J. H. Field and W. E. Boles, Alcheringa 22, 177 (1998). Donahue, J. Archaeol. Sci. 189, 35 (1991). 15. The Genyornis analyzed from MG come from 185 42. A. Baynes, in Taphonomy Symposium 95, Canberra, rior, which in turn placed unprecedented 28–29 April 1995 (Department of Anthropology and stress on the dependent fauna (6, 7, 40). This different sites, from LF 28 sites, and from DM 17 sites. Dromaius from MG were collected from 198 Archaeology, Australian National University, Can- berra, 1995), pp. 12–14. stress, possibly coupled with modest drying sites, from LF 30 sites, and from DM 36 sites. Downloaded from that occurred simultaneously and perhaps 16. Climate data are mean annual temperature and mean 43. D. J. Meltzer and J. I. Mead, in Environments and Extinctions: Man in Late Glacial North America,J.I. some direct human predation, led to annual precipitation calculated with programs de- scribed by McMahon et al. (21). Mead and D. J. Meltzer, Eds. (University of Maine Center for the Study of Early Man., Orono, 1985), pp. megafauna extinction. 17. In Fig. 2 and in general discussions we have converted 145–173. Most of the extinct megafauna were brows- radiocarbon ages to calendar ages using the CALIB 3.0 44. D. L. G. Williams and P. V. Rich, in Vertebrate Pale- ers (1). Vegetation over much of their former Program (45) for samples Ͻ10,000 14C yr B.P., and the equation CalibAge ϭϪ5.85 ϫ ontology of Australasia, P. V. Rich, J. M. Monaghan, range is currently desert scrub and spinifex 10Ϫ6(Conv14CAge)2 ϩ 1.39(Conv14CAge) Ϫ 1807 (E. R. F. Baird, T. H. Rich, Eds. (Pioneer Design Studio, grassland, providing circumstantial evidence in Bard, unpublished data based on paired U-series and Lilydale, Victoria, Australia, 1992), pp. 871–891. 45. M. Stuiver and P. Reimer, Radiocarbon 35, 215 support of major vegetation change. Those taxa 14C dates). However, when citing specific dates, we leave these in radiocarbon years because most are at (1993). with restricted dietary requirements, such as or beyond the limits of radiocarbon dating, for which 46. Supported by the U.S. National Science Foundation, Genyornis, were at greatest risk, whereas more there is no generally accepted terminology or cali- Division of Atmospheric Sciences (G.H.M., B.J.J., and bration curve. All luminescence and U-series dates M.L.F.), the Australian National University (J.W.M., cosmopolitan, opportunistic feeders, such as M.T.M., L.K.A., and G.H.M.), and the University of Dromaius, were able to survive as a new equi- are given in calendar years; the U-series dates have 2␴ uncertainties of about 2%, whereas the lumines- Colorado Faculty Fellowship Program (G.H.M.). We librium was established. Top-level predators cence dates have 1␴ uncertainties of 10 to 15%. thank J. Bowler for initiating our study of the Lake Eyre Basin; A. Baynes, J. Chappell, A. Chivas, P. were unable to sustain themselves in the face of 18. We evaluated the potential of TIMS U-series dating Clark, P. DeDekker, T. Flannery, P. Latz, P. Murray, using 20 Genyornis eggshell fragments of known AAR. reduced prey. M. Smith, and S. Webb for discussions; C. Hart for U-series dates are shown to be reproducible among lab assistance; G. Atkin, O. Miller, and M. Miller for different fragments from the same unit, despite a field assistance; and the station owners and aborig- References and Notes wide range in U concentrations, but some samples inal community leaders in our field areas for land 1. T. F. Flannery, Archaeol. Oceania 25, 45 (1990). have enough common thorium (232Th) to require access and logistical assistance. 2. P. Vickers-Rich, in The Antipodean Ark: Creatures correction for unsupported 230Th. Eggshell has dis- from Prehistoric Australia, S. Hand and M. Archer, Eds. crete pores that allow the developing embryo to (Angus and Robertson, Sydney, 1987), pp. 48–50. exchange gases with the atmosphere, and in some 9 October 1998; accepted 24 November 1998

208 8 JANUARY 1999 VOL 283 SCIENCE www.sciencemag.org