Late Pleistocene human populationbottlenecks, volcanic winter, anddifferentiation of modern humans

The ‘‘Weak Garden of Eden’’ model for the origin and dispersal ofmodern humans (Harpending et al., 1993) posits that modernhumans spread into separate regions from a restricted source, around100 ka (thousand years ago), then passed through population bottle-necks. Around 50 ka, dramatic growth occurred within dispersedpopulations that were genetically isolated from each other. Popula-tion growth began earliest in Africa and later in Eurasia and ishypothesized to have been caused by the invention and spreadof a more efficient Later Stone Age/Upper Paleolithic technology,which developed in equatorial Africa.

Climatic and geological evidence suggest an alternative hypothesisfor Late Pleistocene population bottlenecks and releases. The lastglacial period was preceded by one thousand years of the coldesttemperatures of the Later Pleistocene (271–70 ka), apparentlycaused by the eruption of Toba, Sumatra. Toba was the largestknown explosive eruption of the Quaternary. Toba’s volcanic wintercould have decimated most modern human populations, especiallyoutside of isolated tropical refugia. Release from the bottleneck couldhave occurred either at the end of this hypercold phase, or 10,000years later, at the transition from cold oxygen isotope stage 4 towarmer stage 3. The largest populations surviving through thebottleneck should have been found in the largest tropical refugia, andthus in equatorial Africa. High genetic diversity in modern Africansmay thus reflect a less severe bottleneck rather than earlier populationgrowth.

Volcanic winter may have reduced populations to levels lowenough for founder effects, genetic drift and local adaptations toproduce rapid population differentiation. If Toba caused the bottle-necks, then modern human races may have differentiated abruptly,only 70 thousand years ago.

? 1998 Academic Press Limited

Journal of Human Evolution (1998) 34, 623–651Article No. hu980219

Introduction

Several alternative models for the origin ofmodern human races are currently beingdebated by paleoanthropologists (Aiello,1993; Lahr, 1996). Regional Continuity(Multiregional Evolution) and hybridizationmodels (Afro-European sapiens and Assimi-lation) are based largely on interpretationsof fossil evidence. Three variations on thetheme of a recent African origin for modern

0047–2484/98/060623+29$25.00/0

humans (Replacement, Weak Garden ofEden and Multiple Dispersals) are based oncombinations of evidence from fossils,archaeology, paleobiogeography, paleo-climatology and the genetic structure ofextant human populations. The main pur-pose of this paper is to evaluate and explainone component of the Weak Garden ofEden (GOE) model (Harpending et al.,1993; Sherry et al., 1994; Rogers & Jorde,1995). The Weak GOE assumes a late

? 1998 Academic Press Limited

Stanley H. AmbroseDepartment of Anthropology,University of Illinois,109 Davenport Hall,607 South Mathews Avenue,Urbana, IL 61801, U.S.A.E-mail: Ambrose@uiuc.edu

Received 19 December1996Revision received14 January 1998and accepted 1 February1998

Keywords: human evolution,population bottleneck,founder effect, race, climatechange, volcanic winter,Africa, Palaeolithicarchaeology.

624 . .

Middle Pleistocene African origin for mod-ern humans (Cann et al., 1987; Stonekinget al., 1992), subdivision and dispersalsof populations during the early LatePleistocene, followed by dramatic reduction(bottlenecks) and expansion (releases) inpopulation size in each region. Populationgrowth after the bottlenecks is related to theinvention and spread of modern humantechnology (the Upper Paleolithic or LaterStone Age) during the latter half of theUpper Pleistocene. This paper presents analternative hypothesis: volcanic winterresulting from the super-eruption of Toba(Sumatra) caused the bottleneck andpopulations expanded before the advent ofmodern technology. If this hypothesis iscorrect, modern human races differentiatedabruptly, mainly through the foundereffect, genetic drift and adaptation to localenvironments, only 71–70 ka (thousandyears ago).

Competing models for the origin ofmodern humans

Regional Continuity and Hybridization modelsThe Regional Continuity or MultiregionalEvolution model (Wolpoff, 1989) proposesthat populations of Homo erectus dispersedfrom Africa throughout the Old Worldabout 1·0 m.y.a. and evolved into modernhumans in each region. Gene flow betweenregions prevented speciation and carried allregional populations across the threshold ofmodern humanity in close synchrony, whileallowing regional population characters toevolve. New dates on fossils of Homo erectusfrom Java, southern China and Georgiasuggest hominid dispersal occurred about1·8 m.y.a. (Swisher et al., 1994; Huanget al., 1995; Gabunia & Veuka, 1995; Larick& Ciochon, 1996). If these dates areconfirmed, modern human geographicdifferentiation in the Multiregional modelwould have begun 1·8 m.y.a.

Two variants of the Regional Continuitymodel suggest African immigrants played amore important role in modern human ori-gins outside of Africa, particularly in westernEurasia. The Afro-European sapiens orHybridization and Replacement hypothesis(Bräuer, 1992) posits a large African contri-bution to modern human populations viahybridization with, and assimilation of,Africans in Europe and western Asia. TheAfrican Assimilation model (Smith, 1992)proposes a smaller African genetic contribu-tion to modern human populations inEurope and West Asia, by assimilation ofsmall numbers of African immigrant genes.

Replacement models

Out of Africa in the Later Pleistocene. TheNoah’s Ark (Howells, 1976), Replacement(Stringer & Andrews, 1988), Strong Gardenof Eden (Harpending et al., 1993), AfricanEve (Cann et al., 1987) or Out of Africa(Giles & Ambrose, 1986) model suggeststhat anatomically modern African popula-tions originated sometime in the last 140–200 ka (Cann et al., 1987; Stoneking et al.,1992) and completely replaced archaic onesoutside of Africa sometime during the last100,000 years. Regional variants of thedescendants of Homo erectus, known as‘‘Archaic’’ Homo sapiens (H. heidelbergensis),including their neanderthal descendants inwestern Eurasia, made no genetic contribu-tion to modern humans (Krings et al.,1997). Proponents of Replacement notethat the earliest modern and near-modernhuman fossils occur throughout Africa 100–250 ka, long before they appear elsewhere(Bräuer, 1992; Bräuer et al., 1997; Stringer& Andrews, 1988). The earliest truly ana-tomically modern humans outside of Africaare found on Africa’s doorstep, in theLevant, and date to about 100 ka (Stringer& Andrews, 1988).

The dispersal date of around 100 ka isimportant because it suggests that modernhumans left Africa during oxygen isotope

625

stage 5, the generally warm, humid lastinterglacial period, 2130 to 274 ka (chro-nology of isotope stages after Martinsonet al., 1987). Tchernov (1992a, 1992b) hascompiled an impressive body of paleonto-logical evidence demonstrating that Africanhominids were members of an Afro-Arabianbiotic community that expanded northwardand out of Africa across the Sinai Peninsulato the southern Levant during stage 5. Thiswas only the latest of several such multi-species dispersals out of Africa during warmand humid interglacial phases. During coldglacial periods, including oxygen isotopestages 6 and 4 (2190–130 and 274–60 ka,respectively), cold and arid-adaptedPalearctic biotic communities returned tosouthwest Asia (Tchernov, 1992a, 1992b).

Genetic evidence for a recent African,rather than an ancient Multiregional originfor modern human populations, was firstclearly demonstrated by Lewontin (1972).He showed that total human genetic diver-sity is extremely low (though highest inAfrica) and most is contained within popu-lations in each geographic region. Twenty-five more years of genetic research haveprovided additional support for Lewontin’sobservation that only 210% of human gen-etic variation is accounted for by differencesbetween populations (Relethford, 1995).Humans also have extraordinarily low levelof within- and between-population geneticvariation in comparison to great apes (Ferriset al., 1981; Ruvolo et al., 1993), furthersupporting an extremely recent origin formodern humans.

Mitochondrial DNA base sequence datareinforce the Replacement model. Cannet al. (1987) showed that Africans havemuch greater mtDNA diversity than Euro-pean and Asian populations and inferredthat all non-African populations can betraced back to an African ‘‘Eve’’ who mayhave lived about 140–200 ka (Cann et al.,1987; Vigilant et al., 1991; Horai et al.,1995; Ruvolo et al., 1993; Stoneking et al.,

1992; Stoneking, 1993). Non-African popu-lations have roughly half the genetic diver-sity and thus seem to be no older than halfthis age (assuming amounts of genetic dif-ference are proportional to time), indicatinga more recent origin for non-Africanmodern populations, about 100 ka. Thisinterpretation has been challenged byTempleton (1993, 1996), but improvedstrategies of analysis of large data sets havesurmounted many of these problems (Pennyet al., 1995). Rogers & Jorde (1995) argueTempleton’s method of analysis is unlikelyto work well in populations that have under-gone rapid size expansions. The effects ofdifferent population sizes and amounts ofgene flow on population genetic diversityand on phylogenetic reconstructions posefar more substantial problems for an Africanorigin of modern humans (Relethford,1995; Relethford & Harpending, 1995).However, an African origin for modernhumans receives additional support from Ychromosome (Hammer & Zegura, 1996;Hammer et al., 1997) and nuclear DNAresearch (Wainscoat et al., 1986; Cavalli-Sforza et al., 1988, 1994; Goldstein et al.,1995; Ayala & Escalante, 1996; Ayala,1996a, 1996b; Tishkoff et al., 1996; contraWolpoff, 1996; Templeton, 1996) and isconsistent with the fossil record (see below).

If the Regional Continuity, Afro-European sapiens Hybridization or Assimila-tion models (Wolpoff, 1989; Bräuer, 1992;Smith, 1992) were correct, one would pre-dict at least a few intra-regionally distinctivewidespread alleles in non-African popula-tions that cannot be traced to recentAfrican populations (Stoneking, 1993;Manderscheid & Rogers, 1996; Rogers,1995), especially if regional populationswere established as early as 1·8 m.y.a. Forexample, if neanderthals contributed to thegene pool of modern western Eurasiansthey should now express a unique suite ofdistinctive alleles. However, ancient non-African alleles have not yet been found

626 . .

(Tishkoff et al., 1996; Hammer et al., 1997;Manderscheid & Rogers, 1996) and nean-derthal mtDNA differs substantially fromthat of living humans (Krings et al., 1997).

Craniometric variation, when analyzedin a fashion analogous to that of thegenetic composition of modern humans byLewontin (1972), provides remarkably simi-lar results (Relethford, 1994; Relethford &Harpending, 1994). Craniometric variationwithin geographic populations is far greaterthan between them and differences betweenpopulations are small. Waddle’s (1994)analysis of craniometric variation in fossilhominids also demonstrates a high degreeof similarity among anatomically modernUpper Pleistocene individuals betweenregions, as well as similarity to late MiddlePleistocene Africans, thus supporting anAfrican origin for Upper Pleistocene mod-ern humans. Neanderthals appear to be sub-stantially different in mtDNA (Krings et al.,1997) and morphology from other hominids(Schwartz & Tattersall, 1996), have chrono-logical overlap with modern humans inEurope (Hublin et al., 1996) and show nospecial morphological relationship to west-ern Eurasian modern humans (Waddle,1994; Turbón et al., 1997; Holliday, 1997).If the Multiregional model is correct, thenUpper Pleistocene and modern humansfrom each region should differ from those ofother regions as much as neanderthals differfrom modern humans (Waddle, 1994).Indeed, if southeast Australasian popula-tions left Africa 1·8 m.y.a. (Larick &Ciochon, 1996), they should have evolved amuch more pronounced suite of regionallydistinct morphological characters than thoseof neanderthals, whose ancestors first per-manently occupied western Eurasia only500 ka (Rightmire, 1997). If the UpperPleistocene age of advanced Homo erectusfrom Ngandong, Java, is confirmed (Swisheret al., 1996), it is consistent with Waddle’s(1994) observations and incompatible withRegional Continuity.

The Multiregional model and its variantsappear to be wholly incompatible with fossil,genetic and craniometric evidence describedabove. In the remainder of this paper vari-ants of the Replacement model are consid-ered more accurate and realistic accounts ofthe origin and spread of modern humans.

Replacement models with bottlenecksSeveral studies of nuclear and mitochondrialDNA have identified a significant bottleneckor bottlenecks (reductions in populationsize) followed by substantial populationgrowth (release), sometime during the lastglacial period (Haigh & Maynard Smith,1972; Jones & Rouhani, 1986; Cann et al.,1987; di Rienzo & Wilson, 1991; Harpend-ing et al., 1993; Sherry et al., 1994; Rogers &Jorde, 1995; Takahata, 1993; Takahataet al., 1995; Templeton, 1993; Watsonet al., 1997; Bortoline et al., 1997). Detailedexplanation of the methods by which popu-lation bottlenecks are identified is beyondthe scope of this discussion, but are well-described in the current literature (di Rienzo& Wilson, 1991; Harpending et al., 1993;Marjoram & Donnelly, 1994; Rogers &Harpending, 1992; Rogers, 1995; Rogers& Jorde, 1995; Sherry et al., 1994; Slatkin &Hudson, 1991). The method involves analy-sis of nucleotide sequence differencesbetween pairs of individuals (mismatches)within a population and the frequency dis-tribution of pairwise differences. If a popu-lation has been of relatively constant size fora very long time, then its mismatch distri-bution tends to be multimodal or ragged(Harpending et al., 1993). If a populationhas passed through a bottleneck orexpanded from a small initial size, then itsdistribution tends to be unimodal. In thiscase, the mode of the distribution is oftenapproximately proportional to the time sinceexpansion. The number of mutations withina population increases through time, so themode, mean and maximum number of pair-wise differences increases (Figure 1). If

627

Figure 1. An idealized illustration of mismatch distributions within a population through time. Thenumber of mismatches or pair-wise differences (Ä) between individuals within a population increasesthrough time due to the accumulation of mutations. The frequency distribution of pair-wise differences isstrongly unimodal in populations that have been reduced to a size small enough to eliminate most geneticdiversity. The mode, mean and maximum pairwise difference increases as mutations accumulate throughtime.

mutation rates are known, then the timesince release can be estimated, but anassumption must be made about populationsize through the bottleneck (Relethford &Harpending, 1995). Larger populations con-tain more alleles; therefore higher diversitycould be interpreted as an earlier release.

How small was the bottleneck? Jones &Rouhani (1986) suggest it could reflect atotal Old World breeding population of 40individuals for two centuries or 4000 indi-viduals for 20,000 years. Harpending et al.(1993) provided estimates as low as 500–3000 females. Rogers & Harpending (1992)suggested less than 2000 breeding femalesand Rogers & Jorde (1995) suggests no lessthan 1000. Haigh & Maynard Smith (1972)calculated an effective population size of lessthan 10,000. Ayala & Escalante (1996;

Ayala, 1995, 1996b) suggest population sizenever fell below 4300 individuals. Thehighest estimates of effective population sizeare about 10,000 reproductive females(Takahata et al., 1995; Klein et al., 1993;Erlich et al., 1996; Sherry et al., 1997).Population size was never significantly lowerthan this for an extended period (Takahata,1993). Although an effective population sizeof 10,000 is not considered a bottleneck bysome of these authors, it is very small ifdivided among small demes (Rogers &Jorde, 1995).

Some researchers have suggested a bottle-neck did not occur, based on analyses ofgenes in the Major HistocompatibilityComplex (Klein et al., 1993). MHCpolymorphisms, including the HumanLeukocyte Antigen complex (Ayala, 1995;

628 . .

Ayala & Escalante, 1996) are, however,under strong positive selection because oftheir fundamental role in antigenicity anddisease resistance. They are problematic forreconstructing evolutionary and populationhistory (Erlich et al., 1996; Hickson &Cann, 1997).

Bottlenecks in other species. The genetic struc-ture of eastern African chimpanzee popula-tions (Pan troglodytes schweinfurthii ) (Morinet al., 1994) suggests they survived one ormore bottlenecks in the Late Pleistocene(Rogers & Jorde, 1995; Goldberg, 1996).Several populations that reside within thePleistocene ice age eastern forest refugiumof the montane region centered on westernUganda and eastern Zaïre (Hamilton,1976), have mismatch distributions thatsuggest they experienced a release from abottleneck 267 ka. This date closelycoincides with the start of the last ice age(oxygen isotope stage 4, 70–74 ka). Popula-tions outside of the core Pleistocene refu-gium seem to have been released from abottleneck much more recently, possiblyafter the Last Glacial Maximum (LGM,oxygen isotope stage 2), around 20 ka(Goldberg, 1996). The later release mayalso be evinced by a bimodal distribution ofpairwise differences in mitochondrial DNAsequences in the core refugium populations.The modes are centered on 267 and220 ka (Goldberg, 1996). Bimodal mis-match distributions can be interpreted inother ways (Marjoram & Donnelly, 1994),but the fit of these chimpanzee data withknown environmental history suggests thehypothesis that some populations experi-enced two bottlenecks and releases duringor after the two coldest parts of the lastglacial.

Mismatch analysis of population struc-tures and histories have been determined forfew other species. Cheetahs apparentlyexperienced two severe population bottle-necks in the very recent past (O’Brien et al.,

1987). The earlier one, estimated to haveoccurred 210 ka, may reflect dramaticreduction in the extent of savanna grasslandenvironments at the beginning of the earlyHolocene wet phase (isotope stage 1,212 ka). Because cheetahs are adapted toopen, arid environments, they are unlikely tobe affected by environmental change involv-ing reduction of forests. Their populationhistory is thus expected to out of phase withthat of chimpanzees. I am not aware ofresearch identifying bottlenecks in otherlarge mammals species. Reconstructingpopulation history with mismatch analysisrequires a long, labor-intensive process ofgenerating sequence data on many individ-uals within a population. Goldberg’s (1996)monumental pioneering research on chim-panzees should be emulated with analysisof the genetic structure of other speciesadapted to humid tropical environments.

The Weak Garden of Eden. The Weak GOEmodel (Harpending et al., 1993; Sherryet al., 1994) is a variant of the Replacementmodel that incorporates bottlenecks andreleases during the Upper Pleistocene(Figure 2). The model posits a dispersal ofAfrican modern humans from a restrictedsource into separate regions around 100 ka,but without significant population expan-sion. This was followed by bottlenecks inthe dispersed, subdivided populations. Thecause of bottlenecks is not explained in theWeak GOE. ‘‘Later, starting around 50,000years ago, dramatic population growth andexpansion occurred separately within dis-persed daughter populations that weregenetically isolated from each other’’(Harpending et al., 1993: 495). Populationgrowth after the bottleneck is thought tohave been initiated by the invention of amore efficient Later Stone Age/UpperPaleolithic (LSA/UP) technology in Africa,which increased environmental carryingcapacity (Harpending et al., 1993; Sherryet al., 1994).

629

In this scenario of population history thegeographic locations or sizes of the dis-persed, genetically isolated daughter popu-lations are not clearly specified. This modelimplies initial dispersal and differentiation ofmodern human populations occurred withinAfrica, but after bottlenecks, the later expan-sions of genetically isolated populations areillustrated as having occurred earliest inAfrica and latest in Eurasia (Harpendinget al., 1993, 1996; Sherry et al., 1994),implying population subdivision and disper-sal to other continents may have occurredbefore bottlenecks. If population growth isrelated to the invention and spread ofadvanced technology from a single sourceregion then it seems to imply that at leastsome modern human populations hadalready dispersed from Africa to Eurasiabefore they adopted the new innovations.The Weak GOE model does not posit areplacement of local archaic populations bymodern ones at the time of population

expansion (perhaps with the exception ofthe neanderthals), rather, it suggestsreplacement of earlier Middle Paleolithic byUpper Paleolithic technologies amongpopulations after their dispersal. If the WeakGOE scenario is correct, and if release froma bottleneck occurred earliest in subsaharanAfrica, then the earliest examples of modernhuman technology and modern behaviorshould be found there. Archaeological evi-dence suggests, but does not conclusivelydemonstrate, that the Middle Stone Age(MSA) to Later Stone Age transitionoccurred around 50 ka in East Africa, andthus earlier than the Middle Paleolithic(MP) to Upper Paleolithic transition in westAsia (Ambrose, 1998).

Figure 2. The Weak Garden of Eden model of Harpending et al. (1993; Sherry et al., 1994). Anatomicallymodern humans evolved from a small population in the late Middle Pleistocene, dispersed from the areaof origin, subdivided, and replaced archaic populations in the early Upper Pleistocene. Each populationpassed through a bottleneck after subdivision and/or dispersal. Later, they expand in number following theinvention and diffusion of modern (Later Stone Age and Upper Paleolithic) technology.

Replacement with bottlenecks and MultipleDispersals. The most detailed and cogentmodel of the origin and spread of modernhumans has been proposed by Lahr & Foley(1996; Lahr & Foley, 1994; Foley & Lahr,

630 . .

Figure 3. The Multiple Dispersals model of Lahr & Foley (1994; Lahr, 1996). Modern humanpopulations experienced bottlenecks within Africa during the penultimate glacial (oxygen isotope stage 6,190–130 ka) and at points of dispersal across narrow land bridges and seaways. The first dispersal, tosouthwest Asia around 100 ka, did not result in permanent replacement of Neanderthals. The seconddispersal, to tropical Australasia across the Bab el Mandeb Straits during the first half of the last glacial,and third dispersal, across the Sinai to the Levant around 45 ka, resulted in permanent replacement ofarchaic hominids outside Africa.

1997). The Multiple Dispersals model(Figure 3) proposes that a populationbottleneck occurred during oxygen isotopestage 6, when cold, dry climates causedisolation and differentiation of populationswithin Africa. Additional bottlenecks, fol-lowed by population expansions, could haveoccurred during dispersals from Africathrough physical bottlenecks such as theSinai Peninsula and the Bab el MandebStraits. The first dispersal of anatomicallymodern humans followed the northeastroute out of Africa across the SinaiPeninsula to the Levant 2100 ka, duringwarm, humid oxygen isotope stage 5. Earlymodern human skeletons at the Levantinecave sites of Skhul and Qafzeh and theirassociated Afro-Arabian biota provide thehard evidence for this dispersal event

(Stringer & Andrews, 1988; Tchernov,1992a, 1992b; Bar-Yosef, 1994). At thebeginning of the last ice age the Afro-Arabian biotic community of the southernLevant was abruptly replaced by a Palearcticone, including cold-adapted neanderthals(Tchernov, 1992a, 1992b; Bar-Yosef,1994). Southwest Asia is considered a cul-de-sac in the multiple dispersals model(Lahr, 1996), so these early modernAfricans failed to reach points farther east ornorth. Equipped with an MP technologysimilar to that of the neanderthals, earlymodern humans apparently lacked techno-logical competitive superiority, and wereeither eliminated or pushed south intoAfrica and/or to the southern ArabianPeninsula at the start of the last glacial. Thefailure of early modern humans to survive in

631

the Levant during the early last glacialimplies they were not yet physiologicallyand/or behaviorally well-adapted to coldclimates and Palearctic environments, or atleast not as well-adapted as neanderthals.The first dispersal thus apparently failedto permanently establish modern humansoutside of Africa.

Genetic evidence shows non-Africanpopulations can be divided into southernAustralasian and northern Eurasian popu-lations (Cavalli-Sforza et al., 1988, 1994;Nei & Roychoudhury, 1993). Furtheranalysis suggests these macro-regionalpopulations divided 50–75 ka (Mountain &Cavalli-Sforza, 1997). To explain this deepand early division, Lahr (1996; Lahr &Foley, 1994) proposed the African ancestorsof modern tropical Australasian populationssuccessfully dispersed from Africa via asouthern route before they successfully dis-persed to western and northern Eurasia.Following Kingdon (1993), they posit anIndian Ocean coastal dispersal route totropical Australasia, from the Horn of Africavia the southern Arabian Peninsula, acrossthe Bab el Mandeb Straits at the southernend of the Red Sea. Crossing the Red Seawould have been facilitated by low sea levelsduring the last glacial period, presumablyduring oxygen isotope stage 4 (74–60 ka), orearly stage 3 (60–45 ka). Fossil and archaeo-logical evidence for this dispersal is scarce,and largely relies on evidence for the occu-pation of Australia by 40–60 ka (Lahr &Foley, 1994).

The final dispersal event, at around 45 ka,was again via the northeast corner of Africa,via the Sinai Peninsula to the Levant. Thiscoincided with the transition from theMiddle to the Upper Paleolithic in westernAsia, Europe and southern Siberia. Thisdispersal, by anatomically modern Africanhumans bearing behaviorally modernhuman technology, led to replacement ofneanderthals and other archaic humanpopulations in Europe and higher latitudes

of western Asia. The European and westAsian fossil and archaeological recordsseems consistent with a late dispersaland replacement (Klein, 1989a; 1992,1995). Refined radiocarbon chronologies(Ambrose, 1997, 1998) suggest the transi-tion from the MSA to the LSA occurred250 ka in East Africa; the MP to the UPoccurred 247–43 ka in the Levant (Bar-Yosef et al., 1996) and 43–40 ka in westernEurope and Siberia (Bischoff et al., 1994;Mercier et al., 1995; Goebel & Aksenov,1995; Goebel et al., 1993).

The Multiple Dispersals model cogentlyintegrates the available genetic, fossil andarchaeological evidence with that frompaleobiogeography and paleoclimatology.The explanation for the bottleneck proposedbelow may be more consistent with theWeak GOE model but does not necessarilyinvalidate the Multiple Dispersals scenariounless it critically depends on the timing ofbottlenecks. A scenario of globally synchron-ous bottlenecks is considered below.

What caused Late Pleistocenebottlenecks?

Geneticists have not offered explanations forbottlenecks in dispersed Late Pleistocenepopulations and few outside their field havepaid much attention to this remarkableevent in human population history. Paleoan-thropologists, except for Lahr (1996; Lahr &Foley, 1994) have generally not acknowl-edged its existence and implications, noroffered explanations for its cause. Thehypothesis proposed below (also seeRampino & Ambrose, in press) makes abottleneck seem inevitable. The causes ofbottlenecks and releases are separate issuesbecause these events may have been widelyseparated in time and may have resultedfrom entirely different processes that need tobe considered separately. This unusualseries of events raises several fundamentalquestions:

632 . .

( 1) What caused the bottlenecks? Could ithave been epidemic diseases, globalclimate change, dispersal of smallfounding populations through physicalgeographic bottlenecks, or a globalcatastrophe analogous to the cometaryimpact at the Cretaceous-Tertiaryboundary?

( 2) When did bottlenecks occur? Diddifferent populations experience bot-tlenecks at different times? Did popu-lations experience more than onebottleneck?

( 3) How long did bottlenecks last, andwhen did populations expand? Werereleases soon after bottlenecks oc-curred, or were populations smallfor a very long time? Were releasessynchronous in different populations,or did they occur earlier in some popu-lations and later in others? Was theremore than one release in a population?

( 4) What caused population expansion?Was it the growth of disease-resistantpopulations, exponential growth offounding populations in new environ-ments, global climate change or thespread of technological innovations?

( 5) Where were releases first initiated:sequentially from Africa to the periph-eries of the Old World, simultaneouslyeverywhere, or wherever foundingpopulations were established?

Cause and timing of bottlenecks

Epidemic disease. Were global populationsdecimated by epidemic diseases? Extremelycontagious diseases and/or high populationdensities would be required to sustain trans-mission. Epidemics could only be recog-nized in the fossil record by skeletal pathol-ogies or catastrophic mortality profiles oflarge skeletal populations. This hypothesis iscurrently untestable.

Founder effects. The founder effect in a smalldispersing population crossing barriers to

gene flow, such as the Sinai desert or theBab el Mandeb Straits, could undoubtedlyappear as asynchronous bottlenecks in thegenetic structure of populations in differentregions. This explanation is plausible inLahr’s (1996; Lahr & Foley, 1994) MultipleDispersals scenario.

Global climate change. Global climate changecould have reduced populations during theearly last ice age, oxygen isotope stage 4(74–60 ka) or during stage 6 (190–130 ka).It is notable that the southern Levantinefossil record shows the abrupt replacementof the Afro-Arabian by the Palearcticbiotic community at the stage 5/4 boundary(Tchernov, 1992a, 1992b). The high reso-lution pollen record from Grand Pile,France (Woillard, 1978; Woillard & Mook,1982), shows the rapid onset of cold, drysteppic conditions at 270 ka. Low primaryproductivity during the last ice agecould have significantly reduced humanpopulation size. As noted above, thereplacement of modern humans by neander-thals in the Levant, suggests African modernhumans were rather poorly-adapted to coldclimates.

Volcanic winter. Was a cataclysmic geologicalevent responsible for the bottleneck?Rampino & Self (1992, 1993; Rampinoet al., 1988) have proposed that the super-eruption of Mt Toba, located in northwestSumatra, caused a volcanic winter lastingseveral years. The Toba eruption is dated to273,500&2000 bp by K/Ar (Chesneret al., 1991) and in ice cores it is dated to71 ka (Zeilinski et al., 1996a, 1996b). Tobawas an order of magnitude larger than anyother known Quaternary explosive volcaniceruption (Dawson, 1992; Rose & Chesner,1990; Huff et al., 1992) and is widely con-sidered to have had a significant impact onQuaternary climate, possibly contributing tothe rapid onset and extreme cold of stage 4(Rampino & Self, 1992, 1993).

633

Toba’s volcanic ash can be traced westacross the Indian Ocean and northwestacross India (Rose & Chesner, 1990). Insome Indian Ocean cores the ash is over12 cm thick (Dawson, 1992). It is a wide-spread terrestrial marker bed in India, whereit occurs as primary and reworked airfall ash,in beds that are commonly one to three,and occasionally up to six meters thick(Acharyya & Basu, 1993).

The dense rock equivalent (DRE) volumeof a volcanic ash layer is a standard measureof the magnitude of an explosive eruption.The largest known eruption of the last 454million years is represented by the Ordovi-cian Millbrig Beds in eastern NorthAmerica, which correlate with the BigBentonite of northwestern Europe (Huff

et al., 1992). The Millbrig/Big BentoniteDRE is 1140 km3 of solid rock (Huff et al.,1992). Tambora F5, the largest known his-toric eruption, displaced 20 km3 and Mt StHelens produced 0·2 km3 of volcanic ash(Huff et al., 1992). Toba had a DRE ofabout 800 km3 (Rose & Chesner, 1990). Itwas thus 40 times larger than the largesteruption of the last two centuries and appar-ently the second largest known explosiveeruption in Phanerozoic history.

The high-resolution record of climate andexplosive volcanism in Greenland ice coresclarifies the chronostratigraphic position ofthe Toba super-eruption, its relationship tothe onset of the last ice age and to dramaticclimatic deterioration at 71 ka (Zielinskiet al., 1996a, 1996b; Dansgaard et al.,1993). Oxygen isotope ratios, calcium andsulfur concentrations and electrical conduc-tivity of ice show Toba erupted about 71 ka,between the interstadials (warm periods) ofDansgaard-Oschlger (D-O) events 19 and20 (Zeilinski et al. 1996b). The eruption ismarked by a 6-year period during which thelargest amount of volcanic sulfur of the last110 ka was deposited in the ice core, andmarks the termination of the interstadial ofD-O event 20. This dramatic volcanic event

is followed by 1000 years of the absolutelylowest ice core oxygen isotope ratios of thelast glacial period. In other words, for 1000years immediately following the Toba erup-tion, the earth witnessed temperaturesrelentlessly colder than during the LastGlacial Maximum at 18–21 ka. The first200 years of this stadial event are marked byincreased calcium deposition, indicatinghigh atmospheric dust concentrations, prob-ably from aeolian erosion due to reducedvegetation cover and sediments exposed by adrop in sea level.

The interstadial portion of D-O event 19,which was a warm period that lasted forapproximately 2000 years following thiscoldest millennium, shows that Toba wasnot directly responsible for the onset of thelast glacial, which occurs in the ice core268 ka (Zielinski et al., 1996b). The dis-crepancy in dating of the eruption of Tobaand onset of the last ice age (stage 4)between K/ar (Chesner et al., 1991), deepsea (Martinson et al., 1987) and Greenlandice records (Zeilinski et al., 1996b;Dansgaard et al., 1993) may reflect differ-ences in methods of estimating ages. The icecore dates are estimated by ice depositionrates in ice older than 50 ka, rather thancounting of annual layers.

The global climatic impact of explosivevolcanic eruptions is controlled by severalfactors (Handler, 1989; Sigurdsson, 1990;Rampino & Self, 1988). First, the eruptiveplume must reach a height greater than17 km, into the stratosphere, before volcanicaerosols can be effectively globally redistrib-uted. Several historic eruptions that havealtered global climate had plume heights fargreater than 17 km (Sigurdsson & Carey,1988). Toba’s plume probably reachedtwice this height (Rampino & Self, 1992).Most solar energy falls at low latitudes, soeruptions near the equator cause much moresubstantial cooling due to the reflection ofsolar energy. The reduction in atmosphericvisibility due to volcanic ash and dust

634 . .

particles is relatively short-lived, how-ever, on the order of three to six months(Sigurdsson, 1990). Longer-term globalclimatic cooling is caused by the highreflectance of sulfuric acid haze (Handler,1989). Sulfuric acid molecules stay sus-pended in the upper atmosphere for severalyears.

Ice core evidence provides the ‘‘smokinggun’’ that implicates Toba as cause of thecoldest millennium of the late Pleistocene. Itshows Toba injected more sulfur thatresided in the atmosphere for a longer time(6 years) than any other volcanic eruption inthe last 110 ka (Zielinski et al., 1996a,1996b). These ice core geochemical dataconfirm the original estimates (Rampinoet al., 1988; Rampino & Self, 1992;Sigurdsson, 1990) for the volcanic winterscenario.

If volcanic winter also led to rapid expan-sion of snow cover, it would have alteredalbedo and perpetuated longer-term cool-ing. Ice core oxygen isotope data show thecoldest millennium of the Upper Pleistocenefollowed Toba’s volcanic winter (Zielinskiet al., 1996a, 1996b). Other climatic recordsmay also reflect Toba’s impact. In theGrand Pile pollen record, the Stadial 1 (iso-tope stage 4) environment is described as‘‘particularly rough’’, an adjective usednowhere else in the description of a recordspanning 140 ka (Woillard, 1978). Longpollen sequences from the Bandung area ofJava and marine sediments in Indonesia(van der Kaars, 1991; van der Kaars &Dam, 1995, 1997) and paleobiogeo-graphic evidence indicate large parts ofsoutheast Asia were deforested at this time(Brandon-Jones, 1996; Flenley, 1996).

The climatic effects of volcanic eruptionscan have severe consequences for humanpopulations (Rampino et al., 1988; Handler& Andsager, 1994; Sigurdsson, 1990). Theeffects of Toba’s volcanic winter on terres-trial plant productivity are described ingreater detail elsewhere (Rampino et al.,

1988; Rampino & Ambrose, in press).Tambora caused the year without a summerin 1816 (Harrington, 1992), with snow inJuly and August in New England and globalcrop failures and famines (Sigurdsson &Carey, 1988; Ramaswamy, 1992). TheToba super-eruption is estimated byRampino & Self (1993) to have lowered seasurface temperatures by 3–3·5)C for severalyears. At higher latitudes the cooling effectwould have been substantially amplified.Rampino & Self (1993) estimate a 12)Creduction in summer temperatures at thelatitude of Quebec for two to three yearsfollowing the super-eruption, and longerterm cooling due to increased reflectance ofsolar energy (reduced albedo) caused bygreater snow cover. If Tambora caused theyear without a summer in 1816, Toba couldhave been responsible for six years of relent-less volcanic winter, substantial loweringof plant biomass and disastrous famine(Rampino et al., 1988; Rampino & Self,1992, 1993; Rampino & Ambrose, in press).

Deep sea cores have probably failed toresolve the magnitude of this event and itsimpact on terrestrial environments becausechronological resolution is usually limited tomillennial scale events due to low depositionrates, sediment bioturbation, secondarytransportation of microfossils and thicknessof sampling intervals (Schindel, 1982;Schiffelbein, 1984; Martin, 1993; Kidwell &Flessa, 1995; Roy et al., 1996). Differencesin power to resolve short-term climateevents are well illustrated by comparisons ofGreenland ice core with North Atlanticmicrofossil and isotopic records (Bond et al.,1993; Jouzel, 1994). Only one study of deepsea cores has achieved sub-millennium-tocentury-scale resolution of climate changeby closely-spaced sampling intervals (Bondet al., 1997).

Isolating this millennium in the archaeo-logical record seems highly improbablegiven the difficulties of precisely and accu-rately dating occurrences in this time range.

635

Perhaps rock shelters in the path of the Tobaash plume, northwest of Sumatra (southernIndia and the Arabian Peninsula) preserve alayer of ash from this eruption that could beused as an isochron. Determining the effectof volcanic winter on human morphologywould be even more difficult because of thescarcity of well-dated human fossils. Thepossible effects of large volcanic eruptionson mammalian morphology has only beenwell-documented in a Miocene sequence inArgentina that contains large samples offossils from a continuous sedimentarysequence (Anderson et al., 1995).

Paleobiologists have traditionally focusedon understanding the effects of long termclimate changes and climatic regimes(glacial-interglacial cycles) on biogeography,population history, adaptation and evolution(Potts, 1996; Boaz, 1997). However, short-term, high amplitude climatic variations ondecadal to sub-millennial scales may havehad substantial and immediate impacts onspecies distributions (Roy et al., 1996) andpopulation histories. It would be valuable tofollow the lead of Zielinski et al. (1996a,1996b), who have the luxury of studying anice core record that provides annual resolu-tion. Intensive micro-sampling of strati-graphic sections in deep sea cores, lakes,bogs and varved fluvial sequences for iso-topic, geochemical, botanical, faunal andsedimentological analyses (Burnham, 1993;Webb, 1993; Bond et al., 1997), and iso-topic analysis of incremental structuresin speleothems, tree rings and corals(Schwarcz, 1986; Yapp & Epstein, 1982;Beck et al., 1997) that span the stage 5/4boundary, should be performed in order tofurther assess Toba’s impact on globalclimate and low-latitude environments.

Famine caused by the Toba super-eruption and volcanic winter provides aplausible hypothesis for Late Pleistocenehuman population bottlenecks. Six years ofvolcanic winter, followed by 1000 years ofthe coldest, driest climate of the Late

Quaternary, may have caused low primaryproductivity and famine, and thus may havehad a substantial impact on human popula-tions (Rampino & Ambrose, in press). Manylocal human populations at higher latitudesand in the path of the ash fallout may havebeen eliminated. This may explain theextraordinary magnitude of the bottleneckeffect recorded in our genes. Figure 4 illus-trates a revised Weak Garden of Eden modelthat incorporates both the effect of volcanicwinter and the Multiple dispersals scenarioof Lahr & Foley (1994).

Neanderthals had a more cold-tolerantanatomy (Trinkaus, 1981, 1989; Holliday,1997), and may have been less affected bythis event. The effects of volcanic winter onthem may never be known if they left nomodern descendants. However, ancientneanderthal DNA has now been recovered(Krings et al., 1997). If DNA can beextracted from a large number of neander-thals their population genetic structurecould be examined for bottlenecks bymismatch analysis.

The cause, timing and location of bottleneckreleasesIf population release was due to the naturalincrease (logistic population growth) ofdisease-resistant populations followingepidemics, then growth could have beenrelatively rapid, a function of the intrinsicrate of increase of disease-resistant popula-tions, and the duration of the bottleneckrelatively brief. Its date could have been atany time, but would presumably havebeen relatively soon after the bottleneck.Release could have occurred whereverdisease-resistant individuals survived.

If release was due to natural increase infounder population size after dispersingacross land bridges or narrow straits (Lahr,1996; Lahr & Foley, 1994) then releasedates would vary from 270–50 ka for theearly Australasian dispersal, to 245 ka forthe second Levantine dispersal. In the

636 . .

Figure 4. The Volcanic Winter/Weak Garden of Eden model proposed in this paper. Populationsubdivision due to dispersal within African and to other continents during the early Late Pleistocene isfollowed by bottlenecks caused by volcanic winter, resulting from the eruption of Toba, 71 ka. Thebottleneck may have lasted either 1000 years, during the hyper-cold stadial period between Dansgaard-Oeschlger events 19 and 20, or 10 ka, during oxygen isotope stage 4. Population bottlenecks and releasesare both synchronous. More individuals survived in Africa because tropical refugia were largest there,resulting in greater genetic diversity in Africa.

epidemic and dispersal scenarios the dura-tion of the bottleneck would have been brief.

If a bottleneck occurred during stage 6(Lahr & Foley, 1994), the release woulddate to the onset of warm climate at thebeginning of stage 5, 2130 ka. This date isnot significantly different from the date ofthe origin of modern humans estimatedfrom the coalescence of mtDNA sequencesto ‘‘Eve’’ by Stoneking et al. (1992). Thespecies origin is by definition a bottleneckbecause only one mtDNA sequence survivesfrom that time, though others undoubtedlyexisted. However, the bottlenecks identifiedby Harpending et al. (1993) that requireexplanation occurred long after Eve.

If bottlenecks were caused by the cold,arid climate of isotope stage 4 then theirduration was approximately 10 ka andrelease could have been as late as 60 ka. IfToba alone caused the bottleneck, then

release could have been within a few decadesof volcanic winter at 271 ka. If the bottle-neck persisted for the entire coldest millen-nium following the eruption, then releasewould have coincided with D-O eventinterstadial 19, around 70 ka. In theseclimate-induced bottleneck scenarios thesurviving populations would have expandedsimultaneously when favorable climatesreturned.

Harpending et al. (1993) proposed thatthe release was due to the invention of amore effective UP or LSA technology insubsaharan Africa. Improved technologywould have increased the effectiveness ofhuman adaptations, including access to alarger, more reliable food supply, and wouldhave led to an increase in population size.The earliest LSA industry in Africa appears250 ka at the site of Enkapune Ya Mutorockshelter in the central Rift Valley of

637

Kenya (Ambrose, 1997; Ambrose, 1998).Three sites in Tanzania (Mumba Höle,Nasera and Naisiusiu at Olduvai Gorge)may have a similar antiquity (Mehlman,1989, 1991; Brooks & Robertshaw 1990;Manega, 1993). The MP/UP transitiondates to around 47 ka in the Levant andabout 43 ka in Europe and Siberia(Ambrose, 1997; Bar-Yosef et al., 1996;Marks, 1983; Mercier et al., 1995; Bischoff

et al., 1994; Goebel & Aksenov, 1995;Goebel et al., 1993; Koslowski, 1988). Thechronology of the transition from archaic tomodern human technology thus fits thegenetic evidence for a wave of advancemode of time-transgressive releases fromthe bottleneck in different regionalpopulations (Harpending et al., 1993;Sherry et al., 1994). In other words, releaseearliest in subsaharan African, and latestEuropean and Asian populations in responseto improved technologically-mediatedadaptations.

Is there a clear relationship between therelease from the bottleneck and the originof modern technology? If the bottleneckoccurred 270–60 ka, this does not neces-sarily mean that it caused technological andbehavioral changes that occurred 10–20 kalater. However, Africa apparently had thelargest surviving population through thebottleneck (Relethford & Harpending,1994), probably because of the large size ofits tropical refugia. By analogy with a largerreservoir of genetic diversity, one couldhypothesize that Africans may have main-tained the largest reservoir of accumulatedknowledge on which to base new technologi-cal and social innovations that characterizethe LSA and UP.

Pleistocene population size anddistribution

African archaeological evidenceDoes the archaeological record support thehypothesis that a Pleistocene population

explosion accompanied transition to mod-ern technology (Harpending et al., 1993;Sherry et al., 1994)? The archaeologicalrecord of the early last glacial period insouthern and northern Africa seems to showvery low population densities throughoutmost of the last ice age. In arid regions ofAfrica, MSA/MP and early LSA/UP occur-rences dating to isotope stages 3 and 4(24–74 ka) are remarkably scarce in com-parison to those of the last interglacial (stage5) and the Holocene (stage 1). Caves androckshelter sites that preserve sedimentsdating to both the last interglacial and lastglacial have relatively abundant traces ofoccupation during oxygen isotope stage 5,while last glacial layers have very low den-sities of cultural materials (Butzer, 1988;Klein, 1989b; Deacon, 1995). Sedimenta-tion with sparse intervals of human occu-pation occurs in several rockshelters duringthe last glacial, for example, at BorderCave (Beaumont et al., 1978), Boomplaas(Deacon et al., 1984) and Die Kelders(Grine et al., 1991) in South Africa,suggesting the paucity of human occupationevidence is not an artifact of erosion orsampling bias.

Most MP sites in the Sahara and NorthAfrica appear to date to the last interglacial(Wendorf & Schild, 1992). Only one earlyUP site is known to date to between 40 and30 ka in the Nile Valley (Vermeersch et al.,1984). Sodmein Cave, in the Red SeaMountains of Egypt, also has sparse lastglacial MP and early UP horizons (Van Peeret al., 1996). The Haua Fteah, a huge caveon the coast of Libya, contains a continuoussedimentary record through the last glacial,but with very low densities of MP artefacts(McBurney, 1967). Culturally sterile sedi-ments accumulated between last interglacialand glacial and between MP and early UPoccupations. The earliest UP, which datesto almost 40 ka, is only known fromHaua Fteah and Hagfet et Dabba caves(McBurney, 1967). In the Kenya Rift

638 . .

Valley, the dating of the late MSA remainsuncertain and industries transitional to theLSA are unknown. The early LSA is knownonly from Enkapune Ya Muto, where itapparently began 250 ka (Ambrose, 1998).

Rockshelters in northern Tanzania seemto have more continuous occupationthrough the Upper Pleistocene and transi-tional MSA/LSA industries seem to occuraround 30–50 ka (Mehlman, 1989, 1991;Brooks & Robertshaw, 1990). Continuoussequences of archaeological cultures fromthe late Middle Pleistocene to the earlyHolocene can also be inferred for theSangoan to Lupemban to Tshitolian indus-tries in equatorial central Africa (VanNoten, 1982) and Sangoan to Charaman toBambatan to Tshangulan/Maleme indus-tries in Zimbabwe and southern Zambia(Clark, 1982: 318; Brooks & Roberstshaw,1990; Walker, 1995).

Archaeological evidence thus suggestsAfrican hunter-gatherers fared poorly incold, arid environments at higher latitudesduring the early last ice age, and seem tohave abandoned these regions or lived atpopulation densities below the threshold ofarchaeological visibility (Butzer, 1988).Continuity of lithic technologies within theMSA or from the MSA to the LSA and MPto the UP cannot be demonstrated in morearid parts of southern, northern or easternAfrica. Humans apparently survived withoutsignificant discontinuities in material cul-tural traditions in warmer, wetter equatorialregions, suggesting this region may havebeen a large population reservoir during thelast glacial and volcanic winter. However,there is no evidence for an increase inthe number of archaeological sites duringthe MSA/LSA or MP/UP transitions inAfrica. Therefore the African archaeologicalrecord does not support the hypothesis(Harpending et al. 1993; Sherry et al., 1994)that a Pleistocene population explosionaccompanied transition to modern tech-nology. However, few sites of this time range

have been excavated and they are difficult todate accurately (Ambrose, 1997; Ambrose,1998), so the limited evidence does notactually refute this hypothesis. Improveddating techniques (Renne et al., 1997) thatcan be applied in regions of active volcanismlike East Africa may be useful for furtherevaluating this hypothesis.

Pleistocene population refugiaTropical areas of high rainfall may haveserved as biotic population refugia duringvolcanic winter and isotope stage 4. Severalrefugia have been identified in SoutheastAsia (Brandon-Jones, 1996), including Java,northeastern Indochina (including southernChina) and southern India. Southern Indiashould have been a relatively large tropicalrefugium, but it was effectively buried by ashfrom Toba (Acharyya & Basu, 1993). Onlythe southern tip of the Indian subcontinent(the western Ghats) seems to have served asa refugium for tropical primates during thelast glacial (Brandon-Jones, 1996). Atpresent sea level the available land area insoutheast Asia is small and fragmented intoislands and peninsulas. Although substan-tially larger land areas east, north and southof Toba were exposed during times of lowsea level (the Sunda and Sahul shelves), theywould only have been exposed after theToba super-eruption and onset of isotopestage 4. Reconstructions of paleobiogeogra-phy and paleoclimatology of tropical south-east Asia during the last and penultimateglaciations suggest deforestation and colddry climates (van der Kaars & Dam, 1995;Brandon-Jones, 1996). It thus seemsunlikely that large human populations couldhave survived volcanic winter in tropicalAustralasia.

The continent with the largest land areaover the equator would have maintained thelargest population through the bottleneck.Africa has the largest well-watered land areaover the equator. Three major refuge areasin equatorial Africa have been identified

639

(Hamilton, 1976), located in far WestAfrica, west-central Africa, and the montaneeastern margin of the Congo basin. In eachrefugium annual rainfall is currently greaterthan 2000 mm. Additional refugia may havebeen located along the south coast of SouthAfrica in the vicinity of the Knysna forest,highland Ethiopia, the Nile Valley and theAtlas mountains of NW Africa. Archaeologi-cal evidence discussed above suggests thelargest human populations during thebottleneck probably survived in equatorialAfrica. Indeed, genetic evidence suggestssome equatorial African populations experi-enced population growth around 110 ka,and did not experience a significant bottle-neck at the beginning of the last glacial(Watson et al., 1997).

Population age vs. size through the bottleneck:implications for genetic diversityDoes Africa necessarily have the oldest mod-ern human genome? Africans have the great-est genetic diversity by most measures (butsee Jorde et al., 1995; 1997; Relethford,1997) and this diversity is widely assumed tobe a function of the greater age of Africanpopulations (Cann et al., 1987; Harpendinget al., 1993). Genetic diversity is, how-ever, also a function of population size.Relethford & Harpending (1994, 1995;Relethford, 1995, 1997) note that diversitydue to large population size can be incor-rectly interpreted as being due to greaterpopulation age. This poses problems for thephylogenetic interpretation of the Africanancestry of modern humans (Relethford &Harpending, 1995; Relethford, 1995).Africa has the largest tropical land area overthe equator and would have been the largestrefugium for humans during volcanic win-ter. Africans may have passed through abottleneck with the greatest geneticdiversity simply because of population size.Relethford & Harpending (1994) estimateAfrican population size through the bottle-neck was three times larger than elsewhere.

Large population size may have also permit-ted an earlier bottleneck release and alarger reservoir of behavioral variation andcapacity for innovation.

Human population history during theLater Pleistocene

The pattern of human population changesproposed by Haigh & Maynard Smith(1972) is illustrated in Figure 5(a). Modifi-cations to account for late Quaternaryenvironmental and cultural changes areshown in Figure 5(b). If the bottleneck wascaused by the cold millennium followingvolcanic winter and/or the severe climate ofoxygen isotope stage 4, then the duration ofthe human bottleneck and date of its releaseare likely to have been either 1000 years and70,000 bp, or 10,000 years and 60,000 bp.If mutation rates can be accurately deter-mined, it should be possible to use thegenetic structure of modern human popula-tions to make relatively precise estimates ofpopulation sizes through the bottleneck indifferent regions.

Mismatch data suggest Pleistocene popu-lations were largest in Africa and smallest inAustralasia (Relethford & Harpending,1994, 1995). It is unclear if this revision ofthe Weak GOE invalidates the hypothesis ofearlier releases from bottlenecks in Africadue to improved technology (Harpendinget al., 1993; Sherry et al., 1994). Climaticand technological releases from a bottleneckare unlikely to have occurred at the sametime, but both may have played a role inLate Pleistocene population growth aftervolcanic winter and the onset of the last iceage. Climatic release may have occurredfirst, around either 70 or 60 ka, withtechnologically-aided release trailing at250 ka, beginning in Africa and spreadingto Eurasia by 245–40 ka, as modernhumans and/or modern LSA/UP technologydiffused from the African source area. Is itpossible to recognize a two stage release

640 . .

Figure 5. Effective modern human population size through the late Quaternary. (A) The ‘‘past history ofhuman numbers’’ redrawn from Figure 1 in Haigh & Maynard Smith (1972). The spread of foodproduction is mainly responsible for population increase during the Holocene. (B) Revision of Haigh &Maynard Smith’s estimates of effective population size through time, based on genetic, climatic, andarchaeological evidence discussed in this paper. Population size may have tracked climate change andprimary productivity. Population size was lowest during volcanic winter, but may have remained relativelylow throughout oxygen isotope stage 4. Improved technology and social organization may have permittedincreased population size after 50 ka and buffered the effects of severe climate on population size.

from the bottleneck in human genetic data?The archaeological evidence discussedabove suggests that population densities inmore arid areas of Africa remained relativelylow throughout the last Glacial MSA/MPand early LSA/UP (Butzer, 1988; Wendorf& Schild, 1992). Relative frequencies of lateMSA and early LSA sites in equatorialAfrica do not seem to increase, so a signifi-cant population release due to technology atthe MSA/LSA boundary cannot be demon-

strated. However, chronological resolutionof this crucial time period is very poorbecause there are few reliable techniquesof chronometric dating of this period(Ambrose, 1998) and few sites have beenexcavated.

Anatomically modern humans during theMP/MSA appear to have been less effectivein exploiting resources and responding toenvironmental changes than behaviorallyand technologically modern humans of the

641

UP/LSA (Ambrose & Lorenz, 1990; Klein,1989a, 1989b, 1992). The effects of vol-canic winter and isotope stage 4 on earlyanatomically modern human populationsmay have been much greater than the effectsof the LGM on behaviorally modernhumans.

The last dispersal of modern humansand/or their technology from Africa mayhave occurred during isotope stage 3,247 ka. Although this may have coincidedwith, and may have been facilitated by along, warm interstadial event (D-O intersta-dial 12) (Dansgaard et al., 1993), it was notaccompanied by a significant northwardshift of Afro-Arabian biomes (Tchernov,1992a, 1992b), which suggests modernhumans with advanced LSA/UP technologyhad greater tolerance of cold conditions.Did humans also experience a bottleneckduring the LGM? Modern humans thrivedin the Last Glacial Maximum (isotope stage2) environments of many parts of the OldWorld (Gamble & Soffer, 1990; Soffer &Gamble, 1990), so population densitieswere probably not substantially loweredduring stage 2, despite harsh climaticconditions.

As noted above, chimpanzees apparentlyexperienced a bottleneck at the LGM(Goldberg, 1996). Unless chimpanzeesevolved substantially improved adaptabilityto cold environments between 70 and 20 ka,they may have been much more susceptiblethan modern humans to climate-relatedpopulation fluctuations at the LGM. Theidentification of a bottleneck 267 ka in thegenetic structure of chimpanzee populationsin a Pleistocene core refugium in easternAfrica synchronous with that of humans(Rogers & Jorde, 1995; Goldberg, 1996)suggests Toba’s volcanic winter may haveaffected the population history of manyspecies adapted to warm and humid envi-ronments. Southeast Asian primates appearto have been greatly affected by climatechange at the beginning of the last two

glacial periods (Brandon-Jones, 1996). Thegenetic structure of primate populations inAsian core refugia should also evince asevere bottleneck around 70 ka.

The effect of bottlenecks on modernhuman variation

The parent population of the African mito-chondrial Eve may have lived as late as130–140 ka (Stoneking et al., 1992) and allliving mtDNA lineages can be traced to her.Low genetic diversity in living human popu-lations is thus largely attributable to therecent origins of modern humans. Howmuch of our current low genetic diversitywithin, and difference between populations,can be accounted for by our recent Africanorigin vs. subsequent bottlenecks in dis-persed populations? Alan Rogers (personalcommunication) has addressed this questionby performing computer simulations usinghis MMGEN program to estimate themean number of pairwise differencesexpected under variations of Multiregional,Replacement (Strong GOE) and WeakGOE models, with and without bottlenecks.His results of 1000 iterations of eachscenario are illustrated in Figure 6 andexplained in Appendix 1.

The Multiregional model [Figure 6(a)]produces a global mean of mean pairwisedifferences (MMPWÄ) an order of magni-tude larger than observed in modernhumans (159 in the MR model vs. 212 inextant populations). Recent bottlenecks inthree ancient populations [Figure 6(b)]apparently have little effect on MMPWÄ.However, if only one ancient populationsurvives a bottleneck [Figure 6(c)],MMPWÄ is reduced to 47·2, which is stillfar in excess of that observed in living popu-lations. Our African Eve, who representsthe coalescent of all living mtDNAlineages (Cann et al., 1987), is a bottleneckby definition. Simulating a population his-tory of subdivision around 70 ka without

642 . .

643

bottlenecks, produces a global MMPWÄ of213·4 [Figure 6(d)]. The Weak GOE, witha single bottleneck in Africa before sub-division into three populations, reducesglobal MMPWÄ to 9·3 [Figure 6(e)]. Mod-eling the Weak GOE with bottlenecksoccurring after subdivision is closest to thescenario proposed in this paper and impliedby Harpending et al. (1993). The resultingMMPWÄ of 11·4 [Figure 6(f)] is most simi-lar to that actually observed, but may not besignificantly different from the result ofsimulation of a single bottleneck prior tosubdivision [Figure 6(e)]. Rogers’ simula-tions of Weak GOE scenarios generatesMPWÄ values close to those actuallyobserved in modern humans, but very farfrom those expected in the Multiregionalmodel (Manderscheid & Rogers, 1996).

Total genetic diversity is reduced by abottleneck and diversity within subdividedpopulations is reduced as well. Figure 6(e)shows that if a bottleneck occurs beforesubdivision, each population will share alarge proportion of alleles. However, ifbottlenecks occur after subdivision, eachone should retain a different subset of thepreexisting pool of genetic variants [Figure6(f )]. In this case within-population diver-sity is also reduced, but between groupdifferences should increase. Simultaneousdecrease in total genetic diversity andincrease in between-population differencemay seem counter-intuitive, but can be

easily comprehended by an alternativedescription of the hypothetical Weak GOEscenarios of Figures 6(e) and 6(f). If a singlepopulation has 13 alleles (a-m) and passesthrough a bottleneck before subdivision[Figure 6(e)] the subdivided populationsmay share 8 of 9 alleles (210% difference),because they all lost the same alleles (k-m)through the bottleneck. Differences betweenpopulations are due to post-bottleneck lossor gain of alleles. If bottlenecks occur aftersubdivision [Figure 6(f)], fewer original alle-les may be retained in each population, butall of the original alleles may be presentamong them because the same alleles areunlikely to be eliminated in each population.Between-population difference would in-crease to 230%. The terms diversity anddifference are often used interchangeably,but they should be used to describe totalnumber of alleles and proportions shared,respectively.

The recent African Replacement model,and the Weak GOE with a bottleneck beforesubdivision [Figures 6(d) and 6(e)] bothimply all living human populations shouldstrongly resemble Africans. Bottlenecks aftersubdivision [Figure 6(f)] may resolve thecentral, but rarely stated paradox of theReplacement model presented by the appar-ent magnitude of superficial physical differ-ences between modern populations that arecommonly used to classify us into humanraces: How can modern human races look so

Figure 6. Illustration of Alan Rogers’ (personal communication) simulations of mean pair-wise differencesin mtDNA sequences of human populations for six models of modern human origins. Simulationassumptions and parameters for each scenario are described in Appendix 1. (A) Multiregional Evolution(MR); (B) MR with recent bottlenecks in each region; (C) Ancient African origin with recent Replacementof non-African populations after a bottleneck; (D) Strong Garden of Eden (recent replacement byrecently-evolved Africans) with no bottleneck. The coalescence of mtDNA lineages at the African Eve(2130–140 ka) is a bottleneck by definition because MPWÄ=0 at the species origin; (E) Weak Garden ofEden (recent replacement by Africans) with one bottleneck prior to subdivision and dispersal; (F) WeakGOE with bottlenecks in each population after subdivision and dispersal. Allelic diversity before and afterbottlenecks in E and F is indicated by the letters within rectangles. In E, overlap between populations ishigh but total MPWÄ is low because all populations lost the same alleles (k, l & m) during the bottleneck,before subdivision. In F, each population retained a different subset of the pre-bottleneck gene pool sowithin group diversity is low. However, total MPWÄ is high because all alleles present before thebottleneck are found among the post-bottleneck populations. Gene flow, assumed for all models, isindicated by stippled arrows in A–C but omitted in D–F for clarity of presentation.

644 . .

different if we are all such recent descend-ants of Africans? In other words, if we are allso recently out of Africa, then why don’t weall look more like each other and more likeour African ancestors?

I propose that smaller, isolated Africanpopulations that did survive Toba’s volcanicwinter in refugia within and/or outsideAfrica may have experienced the equivalentof the Founder Effect (Mayr, 1970; Neiet al., 1975; Dobzhansky & Pavlovsky,1953). When populations are reduced tovery small sizes, most genetic diversity islost. A small, random subset of pre-existinggenetic diversity is retained in each isolatedpopulation. If founder populations remainsmall and isolated for many generations,then genetic drift leads to the random loss ofadditional alleles and fixation of others,reducing genetic diversity and increasingbetween-population difference even further.Additional differentiation can occur throughadaptations to local environments. If theToba super-eruption and hyper-cold millen-nium between D-O event interstadials 19and 20 kept human populations small andrestricted to isolated refugia, then these pro-cesses could have operated for about 40generations (assuming 25 years per genera-tion). It is highly improbable for severalisolated populations to retain the samecomplement of alleles through bottlenecks[Figure 6(f)]. Therefore the bottleneckcould have decreased global genetic diversitywhile increasing genetic difference be-tween subdivided populations. Evolutionarychange occurs fastest in small isolatedpopulations. Therefore population differen-tiation may have accelerated around71–70 ka. Without the bottleneck therewould have been less interregional differ-ence and more clinal variation across theOld World. We would all look more likeAfricans. If modern human regional popu-lation differences originated in isolatedregions during the bottleneck then patternsof clinal variation observed in extant

human populations may be due to geneflow after the bottleneck.

Volcanic winter may have inserted a briefPunctuated Equilibrium event (Eldredge &Gould, 1972) (perhaps an exclamationpoint!) in the course of recent human evolu-tionary history, accelerating geographic dif-ferentiation. This bottleneck may resolve theparadox of why modern human populationslook so different from each other yet havesuch a recent African origin.

Summary and conclusions

Geneticists have repeatedly observed thatmodern human populations passed throughvery small population bottlenecks during theearly Upper Pleistocene. A possible causeof bottlenecks among genetically isolatedhuman populations is a six-year long vol-canic winter and subsequent hyper-coldmillennium resulting from the cataclysmicsuper-eruption of Toba, Sumatra, 71 ka.This dramatic climatic event has not yetbeen carefully studied with a diverse rangeof materials and scientific methods; furtherresearch is needed to fully evaluate the mag-nitude of Toba’s volcanic winter on globalclimate and its ecological impacts onhumans and other species.

Population expansion following thebottleneck may have occurred in responseto climatic amelioration. The alternativehypothesis, that population growth wasfueled by improved technology with theinvention of modern human technology andbehavior at the beginning of the LSA/UP inequatorial Africa, can only be confirmed orrejected after much more archaeologicalresearch and an intensive program ofaccurate chronometric dating.

Volcanic winter is hypothesized to haveplayed an important role in human differen-tiation during the Upper Pleistocene. Com-bined with the founder effect, genetic driftand adaptation to local environments, thebottlenecks may account for part of the

645

overall low human genetic diversity, andbetween population differences in the super-ficial physical characters associated withdifferent geographic races. One conse-quence of volcanic winter may have beenrapid differentiation of small, isolatedAfrican immigrant populations into modernhuman races only 71–70 ka. The bottleneckresolves the paradox of the Out of AfricaReplacement model: why do human popu-lations look so different yet have such arecent African origin? When the modernAfrican human diaspora passed through theprism of Toba’s volcanic winter, a rainbowof differences appeared.

Acknowledgements

This paper is dedicated to the late PaulHandler, who stimulated my interest in vol-canic eruptions and climate change. I thankHenry Harpending for explaining mismatchanalysis and the Weak Garden of Edenmodel. I am most grateful to Alan Rogers forhis intense and persistent efforts in review-ing this paper and permitting me to presentthe results of his computer simulations ofmismatch distributions. Tony Goldberg hasprovided immensely valuable explanationsof many aspects of genetic analysis of popu-lation history. Steve Leigh greatly improvedthe manuscript structure and organizationand offered valuable substantive comments,corrections and criticisms. This paperhas also benefited from critical comments,corrections, discussion, and informationprovided by Alan Almquist, EugeneGiles, Marta Lahr, Mike Rampino, JohnRelethford, Olga Soffer, Alan Templeton,Eric Trinkaus, Linda Van Blerkom,Elizabeth Watson, Martin A. J. Williams,Greg Zielinski and two anonymous re-viewers. I take responsibility for remainingerrors of fact or interpretation.

ReferencesAcharyya, S. K. & Basu, P. K. (1993). Toba ash on the

Indian subcontinent and its implications for cor-

relation of Late Pleistocene alluvium. Quatern. Res.40, 10–19.

Aiello, L. C. (1993). The fossil evidence for modernhuman origins in Africa: a revised view. Am. Anthrop.95, 73–96.

Ambrose, S. H. (1997). Implications of geomagneticfield variation for the chronology of the Middle toLater Stone Age and Middle to Upper Paleolithictransitions. J. hum. Evol. 32(4), A2.

Ambrose, S. H. (1998). Chronology of the Later StoneAge and food production in East Africa. J. Archaeol.Sci. 25, 377–392.

Ambrose, S. H. & Lorenz, K. G. (1990). Social andecological models for the Middle Stone Age inSouthern Africa. In (P. Mellars, Ed.) The Emergenceof Modern Humans, pp. 3–33. Edinburgh: EdinburghUniversity Press.

Anderson, D. K., Damuth, J. & Bown, T. M. (1995).Rapid morphological change in Miocene marsupialsand rodents associated with a volcanic catastrophe inArgentina. J. Vert. Paleontol. 15, 640–649.

Ayala, F. J. (1995). The myth of Eve: molecular biologyand human origins. Science 270, 1930–1936.

Ayala, F. J. (1996a). Gene lineages and human evolu-tion. Science 272, 1363–1364.

Ayala, F. J. (1996b). HLA sequence polymorphismsand the origin of humans. Science 274, 1554.

Ayala, F. J. & Escalante, A. A. (1996). The evolution ofhuman populations: a molecular perspective. MolPhylogenetics Evol. 5, 188–201.

Bar-Yosef, O. (1994). The contribution of southwestAsia to the study of modern human origins. In(M. H. Nitecki & D. V. Nitecki, Eds) Origins ofAnatomically Modern Humans, pp. 22–36. New York:Plenum Press.

Bar-Yosef, O., Arnold, M., Mercier, N, Belfer-Cohen,A., Goldberg, P., Housley, R., Laville, H., Meignen,L., Vogel, J. C. & Vandermeersch, B. (1996). Thedating of Upper Paleolithic layers in Kebara Cave,Mt Carmel. J. Archaeol. Sci. 23, 297–306.

Beaumont, P. B., De Villiers, H. & Vogel, J. C. (1978).Modern man in sub-Saharan Africa prior to49,000 bp: a review with particular reference toBorder Cave. S. Afr. J. Sci. 74, 409–419.

Beck, W. J., Récy, J., Taylor, F., Edwards, R. L. &Cabioch, G. (1997). Abrupt changes in earlyHolocene tropical sea surface temperature derivedfrom coral records. Nature 385, 705–707.

Bischoff, J. L., Ludwig, K., Garcia, J. F., Carbonell, E.,Vaquero, M., Stafford, T. W., Jr & Jull, A. J. T.(1994). Dating of the basal Aurignacian sandwich atAbric Romani (Catalunya, Spain) by radiocarbonand uranium-series. J. Archaeol. Sci. 21, 541–551.

Boaz, N. (1997). Eco Homo. New York: BasicBooks.Bond, G., Showers, W., Cheseby, M., Lotti, R.,

Almasi, P., deMenocal, P., Priore, P., Cullen, H.,Hajdas, I. & Bonani, G. (1997). A pervasivemillennial-scale cycle in north Atlantic Holocene andglacial climatic cycles. Science 278, 1257–1266.

Bond, G., Broecker, W., Johnsen, S., McManus, J.,Labeyrie, L., Jouzel, J. & Bonani, G. (1993).Correlations between climate records from North

646 . .

Atlantic sediments and Greenland ice. Nature 365,143–147.

Bortolini, M. C., Zago, M., Salzano, F. M., Silva-Júnior, W. A., Bonatto, S. L., Da Silva, M. C. B. O.& Weimer, T. A. (1997). Evolutionary and anthro-pological implications of mitochondrial DNA vari-ation in African Brazilian populations. Hum. Biol. 69,141–159.

Brandon-Jones, D. (1996). The Asian Colobinae(Mammalia: Cercopithecidae) as indicators ofQuaternary climate change. Biol. J. Linn. Soc. 59,327–350.

Bräuer, G. (1992). Africa’s place in the evolution ofmodern humans. In (G. Bräuer & F. H. Smith, Eds)Continuity or Replacement. Controversies in Homosapiens Evolution, pp. 63–98. Rotterdam: A. A.Balkema.

Bräuer, G., Yokoyama, Y., Falguères, C. & Mbua, E.(1997). Modern human origins backdated. Nature386, 337–338.

Brooks, A. S. & Robertshaw, P. T. (1990). The glacialmaximum in tropical Africa: 22,000–12,000 BP. In(C. Gamble & O. Soffer, Eds) The World at18,000 BP, Vol. 2, Low Latitudes, pp. 121–169.London: Unwin Hyman.

Burnham, R. J. (1993). Time resolution in terrestrialmacrofloras: guidelines from modern accumulations.In (S. M. Kidwell & A. K. Behrensmeyer, Eds)Taphonomic Approaches to Time Resolution in FossilAssemblages, pp. 57–78. Paleontological Society,Short Courses in Paleontology, No. 6. Knoxville:University of Tennessee.

Butzer, K. W. (1988). A ‘‘marginality’’ model toexplain major spatial and temporal gaps in the Oldand New World Pleistocene settlement records.Geoarchaeology 3, 193–203.

Cann, R. L., Stoneking, M. L. & Wilson, A. C. (1987).Mitochondrial DNA and human evolution. Nature325: 31–36.

Cavalli-Sforza, L. L., Menozzi, P. & Piazza, A. (1994).The History and Geography of Human Genes.Princeton: Princeton University Press.

Cavalli-Sforza, L. L., Piazza, A., Menozzi, P. &Mountain, J. (1988). Reconstruction of human evo-lution: bringing together genetic, archaeological andlinguistic data. Proc. natn. Acad. Sci. U.S.A. 85,6002–6006.

Chesner, C. A., Rose, W. I., Deino, A., Drake, R. &Westgate, J. A. (1991). Eruptive history of earth’slargest Quaternary caldera (Toba, Indonesia) clari-fied. Geology 19, 200–203.

Clark, J. D. (1982). The cultures of the MiddlePaleolithic/Middle Stone Age. In (J. D. Clark, Ed.)Cambridge History of Africa, Vol. 1, pp. 248–341.Cambridge: Cambridge University Press.

Dansgaard, W., Johnsen, S. J., Clausen, H. B., Dahl-Jensen, D., Gundestrup, N. S., Hammer, C. U.,Hvidberg, C. S., Steffensen, J. P., Sveinbjornsdottir,A. E., Jouzel, J. & Bond, G. (1993). Evidence forgeneral instability of past climate from a 250-kyrice-core record. Nature 364, 218–220.

Dawson, A. G. (1992). Ice Age Earth. Late QuaternaryGeology and Climate. London: Routledge.

Deacon, H. J. (1995). Two Late Pleistocene-Holocenearchaeological depositories from the Southern Cape,South Africa. S. Afr. Archaeol. Bull. 50, 121–131.

Deacon, H. J., Deacon, J., Scholtz, A., Thackeray,J. F., Brink, J. S. & Vogel, J. C. (1984). Correlationof paleoenvironmental data from the Late Pleis-tocene and Holocene deposits at Boomplaas Cave,southern Cape. In (J. C. Vogel, Ed.) Late CainozoicPaleoclimates of the Southern Hemisphere, pp. 339–352.Rotterdam: A. A. Balkema.

di Rienzo, A. & Wilson, A. C. (1991). Branchingpattern in the evolutionary tree for human mito-chondrial DNA. Proc. natn. Acad. Sci. U.S.A. 88,1597–1601.

Dobzhansky, T. & Pavlovsky, O. (1953). Indeterminateoutcome of certain experiments on Drosophila popu-lations. Evolution 7, 198–210.

Eldredge, N. & Gould, S. J. (1972). Punctuatedequilibria: an alternative to phyletic gradualism. In(T. J. M. Schopf, Ed.) Models in Paleobiology,pp. 82–115. San Francisco: Freeman Cooper.

Erlich, H. A., Bergström, T. F., Stoneking, M. &Gyllensten, U. (1996). HLA sequence polymor-phisms and the origin of humans. Science 274, 1552–1554.

Ferris, S. D., Brown, W. H., Davidson, W. S. &Wilson, A. C. (1981). Extensive polymorphism in themitochondrial DNA of Apes. Proc. natn. Acad. Sci.U.S.A. 78, 6319–6323.

Flenley, J. R. (1996). Problems of the Quaternary onmountains of the Sunda-Sahul region. Quat. Sci.Rev. 15, 557–580.

Foley, R. & Lahr, M. M. (1997). Mode 3 technologiesand the evolution of modern humans. CambridgeArchaeol. J. 7, 3–36.

Gabunia, L. & Veuka, A. (1995). A Plio-Pleistocenehominid from Dmanisi, East Georgia, Caucasus.Nature 373, 509–512.

Gamble, C. & Soffer, O. (1990). The World at18,000 BP, Vol. 2, Low Latitudes. London: UnwinHyman.

Giles, E. & Ambrose, S. H. (1986). Are we all ‘‘out ofAfrica’’? Nature 322, 21–22.

Goebel, T. & Aksenov, M. (1995). Acceleratorradiocarbon dating the initial Upper Paleolithic insoutheast Siberia. Antiquity 69, 349–357.

Goebel, T., Derevianko, A. P. & Petrin, V. T. (1993).Dating the Middle-to-Upper-Paleolithic transition atKara-Bom. Curr. Anthropol. 34, 452–458.

Goldberg, T. L. (1996). Genetics and Biogeography ofEast African Chimpanzees (Pan troglodytes schwein-furthii ). PhD. Dissertation, Cambridge: HarvardUniversity.

Goldstein, D. B., Ruiz Linares, A., Cavalli-Sforza,L. L. & Feldman, M. W. (1995). Genetic absolutedating based on microsatellites and the origin ofmodern humans. Proc. natn. Acad. Sci, U.S.A. 92,6723–6727.

Grine, F. E., Klein, R. G. & Volman, T. P. (1991).Dating, archaeology and human fossils from the

647

Middle Stone Age Levels of Die Kelders, SouthAfrica. J. hum. Evol. 21, 363–395.

Haigh, J. & Maynard Smith, J. (1972). Population sizeand protein variation in man. Genetical Res. 19,73–89.

Hamilton, A. C. (1976). The significance of patterns ofdistribution shown by forest plants and animals forthe reconstruction of Upper Pleistocene paleo-environments: a review. Palaeoecol. Afr. 9, 63–97.

Hammer, M. F., Spurdle, A. B., Karafet, T., Bonner,M. R., Wood, E. T., Novelletto, A., Malaspina, P.,Mitchell, R. J., Horai, S., Jenkins, T. & Zegura, S. L.(1997). The geographic distribution of human Ychromosome variation. Genetics 145, 787–805.

Hammer, M. F. & Zegura, S. L. (1996). The role of theY chromosome in human evolutionary studies. Evol.Anthrop. 5, 116–134.

Handler, P. (1989). The effect of volcanic aerosols onclimate. J. Volcanol. Geotherm. Res. 37, 233–249.

Handler, P. & Andsager, K. (1994). El Nino, volcan-ism, and global climate. Hum. Evol. 22, 37–57.

Harpending, H., Relethford, J. & Sherry, S. T. (1996).Methods and models for understanding humandiversity. In (A. J. Boyce & C. G. N. Mascie-Taylor,Eds) Molecular Biology and Human Diversity,pp. 283–299. Cambridge: Cambridge UniversityPress.

Harpending, H. C., Sherry, S. T., Rogers, A. L. &Stoneking, M. (1993). The genetic structure ofancient human populations. Curr. Anthropol. 34,483–496.

Harrington, C. R. (1992). The Year Without a Summer?Ottawa: Canadian Museum of Nature.

Hickson, R. E. & Cann, R. L. (1997). Mhc allelicdiversity and modern human origins. J. Mol. Evol.45, 589–598.

Holliday, T. W. (1997). Body proportions in LatePleistocene Europe and modern human origins.J. hum. Evol. 32, 423–447.

Horai, S., Hayasaka, K., Kondo, K., Tsugane, K. &Takahata, N. (1995). Recent African origin ofmodern humans revealed by complete sequences ofhominoid mitochondrial DNAs. Proc. natn. Acad.Sci. U.S.A. 92, 532–536.

Howells, W. W. (1976). Explaining modern man:evolutionists versus migrationists. J. hum. Evol. 5,477–495.

Huang, W., Ciochon, R., Gu, Y., Larick, R., Fang, Q.,Schwarcz, H., Yonge, C., de Vos, J. & Rink, W.(1995). Early Homo and associated artifacts fromAsia. Nature 378, 275–278.

Hublin, J-J., Spoor, F., Braun, B., Zonneveld, F. &Condemi, S. (1996). A late neanderthal associatedwith Upper Palaeolithic artefacts. Nature 381, 224–226.

Huff, W. D., Bergström, S. M. & Kolata, D. R. (1992).Gigantic Ordovician volcanic ash fall in NorthAmerica and Europe: biological, tectonomagmatic,and event-stratigraphic significance. Geology 20, 875–878.

Jones, J. S. & Rouhani, S. (1986). How small was thebottleneck? Nature 319, 449–450.

Jorde, L. B., Bamshad, M. J., Watkins, W. S., Zenger,R., Fraley, A. E., Krakowiak, P. A., Carpenter,K. D., Soodyall, H., Jenkins, T. & Rogers, A. R.(1995). Origins and affinities of modern humans: acomparison of mitochondrial and nuclear geneticdata. Am. J. Hum. Genet. 57, 523–538.

Jorde, L. B., Rogers, A. R., Bamshad, M., Watkins, W.S., Krakowiak, P., Sung, S., Kere, J. & Harpending,H. C. (1997). Microsatellite diversity and the demo-graphic history of modern humans. Proc. natn. Acad.Sci. U.S.A. 94, 3100–3103.

Jouzel, J. (1994). Ice cores north and south. Nature372, 612–613.

Kidwell, S. M. & Flessa, K. W. (1995). The quality ofthe fossil record: populations, species, and commu-nities. Ann. Rev. Ecol. Syst. 26, 269–299.

Kingdon, J. (1993). Self-Made Man and his Undoing.London: Simon & Schuster.

Klein, J., Takahata, N. & Ayala, F. J. (1993). MHCpolymorphisms and human origins. Sci. Am. 269(6),78–83.

Klein, R. G. (1989a). The Human Career. Chicago:University of Chicago Press.

Klein, R. G. (1989b). Biological and behavioral per-spectives on modern human origins in southernAfrica. In (P. Mellars & C. B. Stringer, Eds) TheHuman Revolution. Behavioral and biological perspec-tives on the origins of modern humans, pp. 529–546.Edinburgh: Edinburgh University Press.

Klein, R. G. (1992). The archeology of modern humanorigins. Evol. Anthrop. 1, 5–14.

Klein, R. G. (1995). Anatomy, behavior, and modernhuman origins. J. Wld. Prehist. 9, 167–198.

Koslowski, J. K. (1988). Transition from the Middle tothe early Upper Paleolithic in central Europe and theBalkans. In (J. F. Hoffecker & C. A. Wolf, Eds) TheEarly Upper Paleolithic: Evidence from Europe andthe Near East. BAR International Series, 437,pp. 529–546.

Krings, M., Stone, A., Schmitz, R. W., Krainitzki, H.,Stoneking, M. & Pääbo, S. (1997). Neandertal DNAsequences and the origin of modern humans. Cell 90,19–30.

Lahr, M. (1996). The Evolution of ModernHuman Diversity. Cambridge: Cambridge UniversityPress.

Lahr, M. & Foley, R. (1994). Multiple dispersals andmodern human origins. Evolutionary Anthrop. 3,48–60.

Larick, R. & Ciochon, R. L. (1996). The Africanemergence and early Asian dispersals of the genusHomo. Am. Scient. 84, 538–551.

Lewontin, R. C. (1972). The apportionment of humandiversity. Evolutionary Biol. 6, 381–398.

Manderscheid, E. J. & Rogers, A. R. (1996). Geneticadmixture in the Late Pleistocene. Am. J. phys.Anthrop. 100, 1–5.

Manega, P. C. (1993). Geochronology, Geochemistryand Isotopic Study of the Plio-Pleistocene HominidSites and the Ngorongoro Volcanic Highlands inNorthern Tanzania. PhD. dissertation, University of

648 . .

Colorado, Boulder. Ann Arbor: University Micro-films International.

Marjoram, P. & Donnelly, P. (1994). Pairwisecomparisons of mitochondrial DNA sequences insubdivided populations and implications for humanevolution. Genetics 136, 673–683.

Marks, A. E. (1983). The Middle to Upper Paleolithictransition in the Levant. Adv. Wld. Archaeol. 2,51–98.

Martin, R. E. (1993). Time and taphonomy: actualisticevidence for time-averaging of benthic foraminiferal assemblages. In (S. M. Kidwell & A. K.Behrensmeyer, Eds) Taphonomic Approaches to TimeResolution in Fossil Assemblages, pp. 34–56. Paleonto-logical Society, Short Courses in Paleontology,No. 6. Knoxville: University of Tennessee.

Martinson, D. G., Pisias, N. G., Hays, J. D., Imbrie, J.,Moore, T. C., Jr & Shackleton, N. J. (1987). Agedating and the orbital theory of the ice ages: devel-opment of a high-resolution 0 to 300,000-yearchronostratigraphy. Quat. Res. 27, 1–29.

Mayr, E. (1970). Populations, Species, and Evolution.Cambridge: Belknap Press, Harvard University Press.

McBurney, C. B. M. (1967). The Haua Fteah.Cambridge: Cambridge University Press.

Mehlman, M. J. (1989). Late Quaternary Archaeologi-cal Sequences in Northern Tanzania. PhD Disser-tation. Urbana: University of Illinois.

Mehlman, M. J. (1991). Context for the emergenceof modern Man in eastern Africa: some newTanzanian evidence. In (J. D. Clark, Ed.) CulturalBeginnings, pp. 177–196. Bonn: Dr. Rudolf HabeltGMBH.

Mercier, N., Valladas, H. & Valladas, G. (1995). Flintthermoluminescence dates from the CFR laboratoryat GIF: contributions to the study of the MiddlePaleolithic. Quat. Sci. Rev. 14, 351–364.

Morin, P. A., Moore, J. J., Chakraborty, R., Jin, L.,Goodall, J. & Woodruff, D. S. (1994). Kin selection,social structure, gene flow, and the evolution ofchimpanzees. Science 265, 1193–1201.

Mountain, J. L. & Cavalli-Sforza, L. L. (1997). Multi-locus genotypes, a tree of individuals, and humanevolutionary history. Am. J. Hum. Genet. 61, 701–718.

Nei, M., Maruyama, T. & Chakraborty, R. (1975).The bottleneck effect and genetic variability of popu-lations. Evolution 29, 1–10.

Nei, M. & Roychoudhury, A. K. (1993). Evolutionaryrelationships among human populations on a globalscale. Mol. Biol. Evol. 10, 927–943.

O’Brien, S. J., Wildt, D. E., Bush, M., Caro, T. M.,FitzGibbons, C., Aggundey, I. & Leakey, R. E.(1987). East African cheetahs: evidence for twopopulation bottlenecks? Proc. natn. Acad. Sci. U.S.A.84, 508–511.

Penny, D., Steel, M., Waddell, P. J. & Hendy, M. D.(1995). Improved analysis of human mtDNAsequences support a recent African origin for Homosapiens. Mol. Biol. Evol. 12, 863–882.

Potts, R. (1996). Humanity’s Descent. The Consequencesof Ecological Instability. New York: William Morrow.

Ramaswamy, V. (1992). Explosive start to the last iceage. Nature 359, 14.

Rampino, M. R. & Ambrose, S. H. (in press). Volcanicwinter in the Garden of Eden: the Toba super-eruption and the Late Pleistocene human populationcrash. In (G. Heiken & F. McCoy, Eds) Volcanichazards and Disasters in Human Affairs. GeologicalSociety of America Special Publication Series.

Rampino, M. R. & Self, S. (1992). Volcanic winter andaccelerated glaciation following the Toba super-eruption. Nature 359, 50–52.

Rampino, M. R. & Self, S. (1993). Climate-volcanismfeedback and the Toba eruption of 274,000 yearsago. Quat. Res. 40, 269–280.

Rampino, M. R., Self, S. & Stothers, R. B. (1988).Volcanic winters. Ann. Rev. Earth Planet. Sci. 16,73–99.

Renne, P. R., Sharp, W. D., Deino, A. L., Orsi, G. &Civetta, L. (1997). 40Ar/39Ar dating into the histori-cal realm: calibration against Pliny the Younger.Science 277, 1279–1230.

Relethford, J. H. (1994). Craniometric variation inmodern human populations. Am. J. phys. Anthrop.95, 53–62.

Relethford, J. H. (1995). Genetics and modern humanorigins. Evol. Anthrop. 4, 53–63.

Relethford, J. H. (1997). Mutation rate and excessAfrican heterozygosity. Hum. Biol. 69, 785–792.

Relethford, J. H. & Harpending, H. C. (1994). Cranio-metric variation, genetic theory, and modern humanorigins. Am. J. phys. Anthrop. 95, 249–270.

Relethford, J. H. & Harpending, H. C. (1995). Ancientdifferences in population size can mimic a recentAfrican origin of modern humans. Curr. Anthropol.36, 667–674.

Rightmire, G. P. (1997). Deep roots for the Neander-thals. Nature 389, 917–918.

Rogers, A. R. (1992). Errors introduced by the infinitesites model. Mol. Biol. Evol. 9, 1181–1184.

Rogers, A. R. (1995). Genetic evidence for a Pleis-tocene population explosion. Evolution 49, 608–615.

Rogers, A. R. & Harpending, H. C. (1992). Populationgrowth makes waves in the distribution of pairwisegenetic differences. Mol. Biol. Evol. 9, 552–569.

Rogers, A. R. & Jorde, L. (1995). Genetic evidence onmodern human origins. Hum. Biol. 67, 1–36.

Rose, W. I. & Chesner, C. A. (1990). Worldwidedispersal of ash and gases from earth’s largest knowneruption: Toba, Sumatra, 75 ka. Palaeogeog. Palaeo-climato. Palaeoecol. 89, 269–275.

Roy, K., Valentine, J. W., Jablonski, D. & Kidwell,S. M. (1996). Scales of climatic variability and timeaveraging in Pleistocene biotas: implications for ecol-ogy and evolution. TREE 11, 458–463.

Ruvolo, M., Zehr, S., von Dornum, M., Pan, D.,Chang, B. & Lin, J. (1993). Mitochondrial COIIsequences and modern human origins. Mol. Biol.Evol. 10, 1115–1135.

Schiffelbein, P. (1984). Effect of benthic mixing on theinformation content of deep-sea stratigraphicalsignals. Nature 311, 651–653.

649

Schindel, D. E. (1982). Resolution analysis: a newapproach to the gaps in the fossil record. Paleobiol. 8,340–353.

Schwarcz, H. P. (1986). Geochronology and isotopicgeochemistry of speleothem. In (P. Fritz & J-ChFontes, Eds) Handbook of Environmental IsotopeGeochemistry, Vol. 2, The Terrestrial Environment, B,pp. 271–303. Amsterdam: Elsevier.

Schwartz, J. H. & Tattersall, I. (1996). Significance ofsome previously unrecognized apomorphies in thenasal region of Homo neanderthalensis. Proc. natn.Acad. Sci. U.S.A. 93, 10,852–10,854.

Sherry, S., Rogers, A. R., Harpending, H., Soodyall,H., Jenkins, T. & Stoneking, M. (1994). Mismatchdistributions of mtDNA reveal recent human popu-lation expansions. Hum. Biol. 66, 761–775.

Sherry, S. T., Harpending, H. C., Batzer, M. A. &Stoneking, M. (1997). Alu evolution in human popu-lations: Using the coalescent to estimate effectivepopulation size. Genetics 147, 1977–1982.

Sigurdsson, H. (1990). Assessment of the atmosphericimpact of volcanic eruptions. In (V. L. Sharpton &P. D. Ward, Eds) Global Catastrophes in Earth History,pp. 99–110. Geological Society of America SpecialPaper 247.

Sigurdsson, H. & Carey, S. (1988). The far reach ofTambora. Nat. Hist. 97(6), 66–72.

Slatkin, M. & Hudson, R. B. (1991). Pairwise compari-son of mitochondrial DNA sequences in stable andexponentially growing populations. Genetics 129,555–562.

Smith, F. H. (1992). The role of continuity in modernhuman origins. In (G. Bräuer & F. H. Smith, Eds)Continuity or Replacement. Controversies in Homosapiens Evolution, pp. 145–156. Rotterdam: A. A.Balkema.

Soffer, O. & Gamble, C. (1990). The World at18,000 BP, Vol. 1, High Latitudes. London: UnwinHyman.

Stoneking, M. (1993). DNA and recent human evolu-tion. Evol. Anthrop. 2, 60–73.

Stoneking, M., Sherry, S. T., Redd, A. J. & Vigilant, L.(1992). New approaches to dating suggest a recentage for the human mtDNA ancestor. Phil. Trans. R.Soc., Lond. B337, 167–175.

Stringer, C. & Andrews, P. (1988). Genetic and fossilevidence for the evolution of modern humans. Science239, 1263–1268.

Swisher, C. G. III, Curtis, G. H., Jacob, T., Getty, A.G., Suprijo, A. & Widiasmoro (1994). Age of theearliest known hominids in Java. Science 263, 1118–1121.

Swisher, C. G. III, Rink, W. J., Antón, S. C., Schwarcz,H. P., Curtis, G. H., Suprijo, A. & Widiasmoro(1996). Latest Homo erectus of Java: potential con-temporaneity with Homo sapiens in southeast Asia.Science 274, 1870–1874.

Takahata, N. (1993). Allelic genealogy and humanevolution. Mol. Biol. Evol. 10, 2–22.

Takahata, N., Satta, Y. & Klein, J. (1995). Divergencetime and population size in the lineage leading tomodern humans. Theor. Pop. Biol. 48, 198–221.

Templeton, A. R. (1993). The ‘‘Eve’’ hypothesis: agenetic critique and reanalysis. Am. Anthropol. 95,51–72.

Templeton, A. R. (1996). Gene lineages and humanorigins. Science 272, 1363.

Tchernov, E. (1992a). Biochronology, paleoecology,and dispersal events of hominids in the southernLevant. In (T. Akazawa, K. Aoki & T. Kimura, Eds)The Evolution and Dispersal of Modern Humans in Asia,pp. 149–188. Tokyo: Hokusen-Sha.

Tchernov, E. (1992b). Eurasian-African bioticexchanges through the Levantine corridor during theNeogene and Quaternary. In (W. v. Koenigswald &L. Werdelin, Eds) Mammalian Migration and Dis-persal Events in the European Quaternary. Cour.Forsch. Senckenberg 153, 103–123.

Tishkoff, S. A., Dietzsch, E., Speed, Pakstis, A. J.,Kidd, J. R., Cheung, K., Bonné-Tamir, B.,Santachiara-Benerecetti, A. S., Moral, P., Krings,M., Pääbo, S., Watson, E., Risch, N., Jenkins, T. &Kidd, K. K. (1996). Global patterns of linkagedisequilibrium at the CD4 locus and modern humanorigins. Science 271, 1380–1387.

Tishkoff, S. A., Kidd, K. K. & Risch, N. (1996).Interpretations of Multiregional evolution. Science274, 706–707.

Trinkaus, E. (1981). Neanderthal limb proportions andcold adaptations. In (C. B. Stringer, Ed.) Aspects ofHuman Evolution, pp. 187–224. London: Taylor andFrancis.

Trinkaus, E. (1989). The Upper Pleistocene transition.In (E. Trinkaus, Ed.) The Emergence of ModernHumans, pp. 42–66. Cambridge: CambridgeUniversity Press.

Turbón, D., Pérez-Pérez, A. & Stringer, C. B. (1997).A multivariate analysis of Pleistocene hominids: test-ing hypotheses of European origins. J. hum. Evol. 32,449–468.

van der Kaars, W. A. (1991). Palynology of easternIndonesian marine piston cores: a Late Quaternaryvegetational and climatic record for Australasia.Palaeogeog. Palaeoclimatol. Palaeoecol. 85, 239–302.

van der Kaars, W. A. & Dam, M. A. C. (1995). A135,000-year record of vegetational and climaticchange from the Bandung area, west Java, Indonesia.Palaeogeog. Palaeoclimatol. Palaeoecol. 117, 55–72.

van der Kaars, W. A. & Dam, M. A. C. (1997).Vegetation and climate change in west-Java,Indonesia, during the last 135,000 years. Quat. Int.37, 73–80.

Van Peer, P., Vermeersch, P. M., Moyersons, J. & VanNeer, W. (1996). Palaeolithic sequence of SodmeinCave, Red Sea Mountains, Egypt. In (G. Pwiti & R.Soper, Eds) Aspects of African Archaeology, pp. 149–156. Harare: University of Zimbabwe Publications.

Van Noten, F. (1982). The Archaeology of Central Africa.Graz: Akademische Druck.

Vermeersch, P. M, Paulissen, E., Gijselings, G., Otte,M., Thomas, A., Van Peer, P. & Lauwers, R. (1984).33,000-yr old chert mining site and related Homo inthe Egyptian Nile Valley. Nature 309, 342–344.

650 . .

Vigilant, L., Stoneking, M., Harpending, H., Hawkes,K. & Wilson, A. C. (1991). African populations andthe evolution of human mitochondrial DNA. Science253, 1503–1508.

Waddle, D. (1994). Matrix correlation tests supporta single origin for modern humans. Nature 368,452–454.

Wainscoat, J. S., Hill, A. V. S., Boyce, A. L., Flint, J.,Hernandez, M., Thein, S. L., Old, J. M., Lynch,J. R., Falusi, A. G., Weatherall, D. J. & Clegg, J. B.(1986). Evolutionary relationships of humanpopulations from an analysis of nuclear DNApolymorphisms. Nature 319, 419–493.

Walker, N. J. (1995). Late Pleistocene and Holocenehunter-gatherers of the Matopos: an archaeologicalstudy of change and continuity in Zimbabwe.Studies in African Archaeology 10. Uppsala: SocietasArchaeologica Uppsala.

Watson, E., Forster, P., Richards, M. & Bandelt,H.-J. (1997). Mitochondrial footprints of humanexpansions in Africa. Am. J. Hum. Genet. 61, 691–704.

Webb, T., III. (1993). Constructing the past fromLate-Quaternary pollen data: temporal resolutionand a zoom lens space-time perspective. In (S. M.Kidwell & A. K. Behrensmeyer, Eds) TaphonomicApproaches to Time Resolution in Fossil Assemblages,pp. 79–101. Paleontological Society, Short Coursesin Paleontology, No. 6. Knoxville: University ofTennessee.

Wendorf, F. & Schild, R. (1992). The Middle Paleo-lithic of North Africa: a status report. In (F. Klees &R. Kuper, Eds) New Light on the Northeast AfricanPast, pp. 40–78. Africa Praehistorica 5. Köln:Heinrich-Barth-Institut.

Woillard, G. M. (1978). Grand Pile peat bog: a con-tinuous pollen record for the last 140,000 years.Quat. Res. 9, 1–21.

Woillard, G. M. & Mook, W. G. (1982). Carbon-14dates at Grande Pile: correlation of land and seachronologies. Science 215, 159–161.

Wolpoff, M. (1989). Multiregional evolution: the fossilalternative to Eden. In (P. Mellars & C. Stringer,Eds) The Human Revolution, pp. 62–108. Edinburgh:Edinburgh University Press.

Wolpoff, M. (1996). Interpretations of Multiregionalevolution. Science 274, 704–706.

Yapp, C. J. & Epstein, S. (1982). Climatic implicationsof D/H ratios of meteoric water over North America(9500–22,000 BP) as inferred from ancient woodcellulose C-H hydrogen. Earth Planet. Sci. Lett. 34,333–350.

Zielinski, G., Mayewski, P. A., Meeker, L. D.,Whitlow, S. & Twickler, M. S. (1996a). A110,000-yr record of explosive volcanism from theGISP2 (Greenland) ice core. Quat. Res. 45, 109–118.

Zielinski, G., Mayewski, P. A., Meeker, L. D.,Whitlow, S. & Twickler, M. S. (1996b). Potentialimpact of the Toba mega-eruption 271,000 yearsago. Geophys. Res. Lett. 23, 837–840.

Appendix 1

Parameters for simulations of mean pair-wise differences (MPWÄ) in mtDNAsequences of human populations under dif-ferent models of modern human origins(Figure 6) generated by Alan Rogers (per-sonal communication). Simulations wererun using MMGEN version 3·7 written byRogers (see Rogers, 1992, 1995; Rogers &Harpending, 1992; Rogers & Jorde, 1995;Manderscheid & Rogers, 1996; Harpendinget al., 1993; Sherry et al., 1994). Pairwisedistributions were generated from a sampleof 50 individuals in each subpopulation, soMPWÄ for each iteration is calculated on asample of pairwise comparisons between150 individuals. The formula for number ofpairwise comparisons is

n(n"1)/2, (1)

so for 50 individuals there are 1225 pairwisecomparisons. Simulations of mean pairwisedifferences were repeated 1000 times foreach model and the mean of mean pairwisedifferences (MMPWÄ) for the global popu-lation was calculated. The mutation modelused by Rogers assumes a finite number ofsites in which mutations can occur (finite-flat, sites=360). Model F is most similar tothe scenario of population history proposedby Harpending et al. (1993) and producesan MMPWÄ closely similar to that observedin modern humans (212).

The mathematic definitions of the par-ameters used can be obtained in the refer-ences cited above. The two most importantparameters (quoted from Sherry et al., 1994:762) ‘‘are

è=2 Nu, (2)

the conventional product of female effectivepopulation size N and the locus-specificmutation rate u (per generation); and

ô=2 tu (3)

651

Appendix

Model è mn ô K MMPWÄ

A. Multiregional evolution (MR),with no bottlenecks

1000·0 0·1 infinite 3 158·61

B. Multiregional, with recentbottlenecks in three populations

1000·00·1

1000·0

0·10·00·1

7·00·1

infinite

333

124·26

C. Ancient African Eve, with a recent bottleneckand extinction of two non-African lineages

1000·00·1

1000·0

0·10·00·1

7·00·1

infinite

313

47·23

D. Strong Garden of Eden (recent origin and replacement).No bottleneck after origin of H. sapiens.The coalescent (Eve) is a bottleneck

1000·01·0

0·10·0

13·0infinite

31

13·56

E. Weak Garden of Eden, with a recentbottleneck in Africa before subdivision into three populations

1000·00·1

1000·01·0

0·10·00·10·0

7·00·15·9

infinite

3131

9·26

F. Weak Garden of Eden, with recentbottlenecks after subdivision into three populations

1000·00·1

1000·01·0

0·10·00·10·0

7·00·15·9

infinite

3331

11·73

the date of expansion measured in unitsof mutational time, where t is time ingenerations.’’ Generation length is 25years.

The table below describes the conditionsfor each scenario illustrated in Figure 6.Gene flow (mn) is number of individuals pergeneration and K is the number of popula-tions surviving during time ô. There is nogene flow during bottlenecks. The simula-tions do not specify the age of Eve. In eachsimulation, Eve’s age is a random variable,

and Eve’s age is likely to be different in eachsimulation. In models D–F, Eve’s popula-tion grows from a small size (è=1) at ô=13before present. Assuming mutation ratesdetermined by Stoneking et al. (1992), ô=13is equivalent to 2130–140 ka. Bottlenecksare set to begin at ô=5·9 after Eve (271 ka),to persist for ô=0·1 (21000 yr) and to endat ô=7 before present (270 ka). The firstline of each model’s parameters is the mostrecent set of conditions in the populationhistory scenario.