Variability selection in hominid evolution

Evolutionary Anthropology 81 ARTI CLES This is a corrected version of an article that was published in Vol. 7, No. 3 of Evolutionary Anthropology. In the original version, text was omitted from the end of the first paragraph on p. 94. Please refer to this version for all future uses. The original citation information remains (Evol Anthropol 7:81–96, 1998). Variability Selection in Hominid Evolution RICHARD POTTS Variability selection (abbreviated as VS) is a process considered to link adaptive change to large degrees of environment variability. Its application to hominid evolution is based, in part, on the pronounced rise in environmental remodeling that took place over the past several million years. The VS hypothesis differs from prior views of hominid evolution, which stress the consistent selective effects associated with specific habitats or directional trends (e.g., woodland, savanna expansion, cooling). According to the VS hypothesis, wide fluctuations over time created a growing disparity in adaptive conditions. Inconsistency in selection eventually caused habitat-specific adaptations to be replaced by structures and behaviors responsive to complex environmental change. Key hominid adaptations, in fact, emerged during times of heightened variability. Early bipedality, encephalized brains, and complex human sociality appear to signify a sequence of VS adaptations—i.e., a ratcheting up of versatility and responsiveness to novel environments experienced over the past 6 million years. The adaptive results of VS cannot be extrapolated from selection within a single environmental shift or relatively stable habitat. If some complex traits indeed require disparities in adaptive setting (and relative fitness) in order to evolve, the VS idea counters the prevailing view that adaptive change necessitates long-term, directional consistency in selection. NATURAL SELECTION AND HOMINID EVOLUTION Natu ral selection is th e process by wh ich adaptive stru ctu res are evolved an d m ain tain ed. As a resu lt of th is cen tu ry’s u n ion of popu lation biology, gen etics, an d paleon tology (th e n eoDarwin ian syn th esis), n atu ral selection is regarded as th e m ain cau se of ch an ge in organ ism s in relation to th eir su rrou n din gs.1–3 Applyin g th is prin ciple to h u m an s, paleoan th ropologists h ave looked to n atu ral selection th eory to accou n t for m an y aspects of Richard Potts conducts excavations at Olorgesailie and Kanam, Kenya and at Bose, China. His research at these sites and previous work at Olduvai Gorge examines the behavior of early hominids and ecological influences on evolution. He is director of the Smithsonian’s Human Origins Program, National Museum of Natural History, Washington, DC. E-mail: [email protected] Key words: variability selection; hominids; environment; adaptation; natural selection; evolution. h om in id evolu tion . Recen t advan ces in en viron m en tal scien ces m ake it in creasin gly possible, h owever, to assess directly th e adaptive settin gs of early h u m an s. Th is approach en ables a m ore exact evalu ation of h ow en viron m en tal ch allen ges, an d th u s n atu ral selection , affected early h om in ids. Usin g en viron m en tal records, m y pu rpose h ere is to advan ce a n ew idea abou t h ow selection in flu en ced h om in id evolu tion an d resu lted in com plex adaptive ch an ge. Natu ral selection depen ds on gen etic variation s, wh ich origin ate idiosyn cratically by m u tation an d recom bin ation . Th e varian ts m ay resu lt in d iffer en t d evelop m en ta l p a th wa ys, ph en otypes, an d ways of in teractin g with en viron m en ts. Natu ral selection is th e biased reprodu ction of certain of th ese gen etic altern atives based on an organ ism ’s ch aracteristics an d in teraction s. Lewon tin 4 defin ed th ree n ecessary an d sufficien t con dition s for n atu r a l selection to op er a te: (1) th e in dividu als or u n its of selection m u st vary (e.g., gen etic variation ); (2) som e in dividu als m u st be m ore fit th an oth ers (differen tial fitn ess); an d (3) th ere m u st be a correlation between th e fitn ess of paren ts an d offsprin g (h eritability). Man y biologists parse th is th ird con dition in to two, addin g th e im portan t proviso of selective con sisten cy from on e gen eration to th e n ext: Th u s (3a) th ere m u st be a correlation between paren ts an d offsprin g in m orp h ology, p h ysiology, a n d b eh a vior (ch aracters are h eritable), an d (3b) th ere m u st be a con sisten t in teraction between th e organ ism an d its en viron m en t du e to th e in h erited ch aracters; th is in teraction determ in es fitn ess.5 Selective con sisten cy, or lon g-term , u n iform selection pressu re, is largely assu m ed to be th e way by wh ich adaptive com plexity evolves.6-8 Th e con son an t effects of n atu ral selection , gen er a tion a fter gen er a tion , r esu lt in en viron m en t-specific adaptation .9 As m an ifested in th e stu dy of h u m an evolu tion , th is paradigm im plies th at early h om in ids were ‘‘adapted to’’ a p a r ticu la r a n cestr a l h a b ita t, som etim es term ed th e en viron m en t of evolu tion ary adaptedn ess.10,11 As I will explain , h owever, adaptive con dition s over th e span of h om in id evolu tion were h igh ly in con sisten t on a local to global scale. Th is h as im portan t im plication s for ou r u n derstan din g of h om in id adaptation . ENVIRONMENTAL HYPOTHESES OF HOMINID EVOLUTION Cen tral to ou r con cern with n atu ral selection is th e apparen t fit between organ ism s an d th eir en viron m en ts. Accordin g to both advocates 3,8 an d critics 12,13 of adaptation ist th in kin g, th e extern al world sets certain problem s th at organ ism s solve. Natu ral selec- 82 Evolutionary Anthropology tion is th e process th at creates th ese solu tion s. It is a process ‘‘by wh ich th e organ ism provides a better an d better ‘solu tion ’ to th e ‘problem ’, an d th e en d resu lt is th e state of bein g adapted.’’12 Applyin g th is th in kin g to h om in id evolu tion leads to certain qu estion s: Wh a t wa s th e exter n a l wor ld of h om in ids like? Wh at problem s did it pose? Paleoan th ropologists seek to u n derstan d wh y bipedal walkin g, m akin g ston e tools, an d growin g large brain s were ben eficial. Th e key qu estion is: Wh at adaptive problem s were solved by th ese im portan t developm en ts? Virtu ally all en viron m en tal h ypoth eses of h om in id evolu tion an swer th is qu estion by focu sin g on a specific type of h abitat. Tradition ally, m ost research ers h ave followed th e savan n a h ypoth esis.7,14–18 Th is states th at adaptation to drier, cooler, an d m ore open h abitats was th e h allm ark of h om in id evolu tion from th e late Miocen e th rou gh th e Pleistocen e. Over th e past decade, p a leon tologists a n d p a leoclim a tologists h ave looked to deep-sea records to sh ow th at m arked global coolin g an d dryin g did occu r, m ain ly in a steplike m an n er. Th e overall tren d is th ou gh t to h ave favored th e evolu tion of a r id -a d a p ted ta xa , in clu d in g h om in ids in Africa.19–23 Th e origin of bipedality, ston e toolm akin g, in creased brain size, oth er adaptive tren ds in Hom o, an d key tu rn over even ts (species origin ation s an d extin ction s) h ave th u s been lin ked specifically to coolin g, dryin g, an d savan n a expan sion — th at is, to a stron g expression of th e glob a l tr en d in Afr ica a n d E u r a sia.18,21,24,25 Yet th e savan n a h ypoth esis h as been ch allen ged. Data from som e h om in id sites suggest that the shift to open, grassdom in ated en vir on m en ts h a p p en ed later th an previou sly th ou gh t.26–29 In itia l p a leob ota n ica l a n d fa u n a l eviden ce from Aram is, E th iopia (ca. 4.4 Ma), su ggests th at th e m ost prim itive h om in id , Ardipith ecu s ram idu s, 30,31 was associated with a relatively closed, tree-dom in ated h abitat.32 An atom ical eviden ce from Sterkfon tein , Sou th Africa (ca. 3.5 to 2.5 Ma), is th ou gh t to im ply th at Au stralopithecu s african u s retain ed th e fu n ction al capacity to clim b trees an d was, in fact, adapted to forested con dition s.33,34 On th e basis of fossil p ollen d a ta fr om ARTI CLES Makapan sgat, Sou th Africa (ca. 3.3 to 3.0 Ma), Rayn er an d colleagu es 35 h ave cou n tered th e tradition al savan n a h ypoth esis with a ‘‘forest h ypoth esis,’’ wh ich states th at closed vegetation con dition s played an im portan t role in th e adaptive evolu tion of early bipeds. Com parin g forest an d open woodlan d ch im pan zees, Boesch -Ach erm an n an d Boesch 36 argu e, m oreover, th at key h om in id traits—tool u se, h u n tin g, social cooperation , an d food sh arin g— are m an ifested m ost often in forested h abitats. Th u s, th ese research ers favor replacin g th e savan n a m odel with a forest adaptation m odel of early h om in id evolu tion . Fin ally, based on a record of season al scaven gin g opportu n ities in r ip a r ia n wood la n d s of th e m od er n Ser en geti, Blu m en sch in e 37 an d Cavallo an d Blu m en sch in e 38 h ave . . . bipedality, toolmaking, meat eating, brain expansion, and other important hominid traits are usually considered to have evolved in specific types of environment and response to the problems they posed. proposed th at early m eat- or m arroweatin g toolm akers preferen tially exploited wooded zon es with in th e savan n a m osaic of E ast Africa. An in itial isotopic stu dy29 sh ows th at h om in id artifacts from basal Bed II Oldu vai wer e d ep osited in a C 3 -d om in a ted zon e, a grassy woodlan d, apparen tly addin g su pport to th e woodlan d m odel in on e level at Oldu vai. Sa va n n a , wood la n d , gla cia l, a n d oth er h abitat-specific h ypoth eses tie in well with th e prevailin g view th at com p lex a d a p ta tion s evolve in r espon se to con sisten t en viron m en tal situ ation s or tren ds, an d depen d on con son an t selective resu lts over m an y gen eration s. Th u s bipedality, toolm akin g, m eat eatin g, brain expan sion , an d oth er im portan t h om in id traits are u su ally con sidered to h ave evolved in specific types of en viron m en t an d respon se to th e problem s th ey posed. AN ALTERNATIVE VIEW OF ADAPTIVE CHANGE A differen t view of h om in id evolu tion stem s from th e fact th at a popu lation m ay face m an y rem odelin gs of its h abitat over geologic tim e. It m ay th u s evolve in respon se to a wide ran ge of adaptive con dition s, expressed over periods far lon ger th an a gen eration . Th is view stresses th e in con sisten cy of selection over lon g tim e span s, an d th u s departs from th e prevailin g paradigm of adaptive evolu tion via lon gterm direction al selection .39,40 E n viron m en tal records worldwide in dicate th at flu ctu ation s grew m ore an d m ore extrem e from th e Miocen e to Recen t. E xtrem es su ch as glacialin terglacial an d arid-m oist cycles were m an ifested over lon ger an d lon ger in tervals, u p to 100,000 years or m ore du rin g th e last h alf of th e Pleistocen e. I su ggest, th erefore, th at sign ifican t ch an ges du rin g h om in id evolu tion resu lted n eith er from statistically predictable effects of n atu ral selection over m an y gen eration s n or from adaptation to an y on e h abitat type or gen eral en viron m en tal tren d. We m ay th u s con sider two ideas poten tially relevan t to h om in id evolu tion : First, en viron m en tal variability cau sed stron g disparities in adaptive settin gs over tim e; secon d, th e resu ltin g in con sisten cies in n atu ral selection h ad a powerfu l in flu en ce on evolu tion ary ch an ge. Th e followin g section s exam in e th ese two ideas. EVIDENCE OF ENVIRONMENTAL VARIABILITY Th e isotopic record from deep-sea cores (Fig. 1a) sh ows m arked oxygen (d18 O) en rich m en t from six m illion years ago to th e presen t. Th is tren d in dicates th e presen ce of a m arked global coolin g tren d associated with th e growth of glaciers on lan d. Coolin g was accom pan ied by con tem poran eou s dryin g. Th e overall clim ate cu rve was com posed, h owever, of m an y in tricate reversals in th e tren d. Th e isotope data, in fact, sh ow a two-to-th ree-fold in crease in th e am plitu de of flu ctu ation over th e span of h om in id evolu tion . Evolutionary Anthropology 83 ARTI CLES Figure 1. (A) Oxygen isotope curve for the past six million years, based on composite data on benthic foraminifera from deep-sea cores.95 The measurement, d18O (a ratio between 18O and 16O), is a proxy of global glacial ice volume and related temperature change. The curve shows an overall trend toward cooling and greater ice volume, especially since ca. 2.8 Ma. The trend was composed, though, of many deviations, which became larger over time. (B) d18O variation from the late Oligocene to the present. The total range of variation in d18O is shown per million years from 27 Ma to the present. Variation of 0.3 to 0.5 parts per mil (ppm) is evident for most of the Neogene. At 6 Ma, however, climatic variability rose sharply. The variability continued to increase and rose to 1.9 ppm during the past 1 myr. Although the potential for diagenetic alteration and damping of the isotopic signal increases with age, the published data have been screened for diagenetic effects and are considered to reflect the original oceanic water d18O composition and temperature. Data from references Prentice and Denton,96 Woodruff and coworkers,97 Wright and Miller,98 and Miller and Moutain.99 (C) Climatic variation from 900 ka to present, based on pollen from the Tenaghi Phillipon peat bog, Macedonia (time scale approximate).42 The climate index is a summation of total tree (T), pine (P), and oak (Q) pollen: CI 5 T 1 2P 1 4Q, where CI 5 0 indicates arid, treeless steppe; CI 5 400 indicates moist temperate forest. The range of variability becomes larger around 700 ka, approximately the time of increased variability in the global oxygen isotope curve. Ocean cores also record th e presen ce of win d-blown con tin en tal du st, p r oxies for r a in fa ll a n d vegeta tion cover on lan d. Su ch m easu res, it is argu ed,21 poin t to a distin ct dryin g tren d in Africa. Th e data, in fact, dem on strate th at, over tim e, lon ger an d m ore extrem e cycles of clim atic oscillation were added to th e preexistin g variability, especially at 2.8 Ma, 1.7 Ma, an d 1.0 Ma.21,41 Accordin g to isotopic an d du st records, in creased flu ctu ation was as m u ch a h allm ark of global en viron m en t after th e Miocen e as was th e coolin g-aridity tren d. Oxygen isotope m easu rem en ts back to th e late Oligocen e (Fig. 1b) stron gly su ggest th at h om in id evolu tion coin cided with th e largest recorded clim atic oscillation s of th e late Cen ozoic. Th e variability eviden t in deep-sea 84 Evolutionary Anthropology records is paralleled by an equ ally com plex h istory of rain fall, tem peratu re, an d vegetation on lan d. Based on E u ropean pollen sequ en ces (Fig. 1c) an d altern atin g layers of bu ried soil an d loess (win d-blown silt) in Asia an d eastern E u rope, sh ifts between den se, m oist forest an d cold, dry steppe occu rred repeatedly du rin g th e past 1 m illion years over large region s.42–45 Th e size of African lakes ch an ged dram atically over th e last glacial cycle, rou gh ly coin cidin g with th e sh rin kage an d expan sion of tropical forest.46 Organ ic-rich layers kn own as sapropels, wh ich were exu ded in to th e Mediterran ean Sea by th e Nile, fu rth er su ggest th at sh arp oscillation s in precipitation a n d vegeta tion p r eva iled over th e n orth eastern qu arter of Africa du rin g th e past h alf m illion years.47 Paleoclim ate an alysis h as bu rgeon ed in recen t years an d h as docu m en ted th e n early con tin u ou s spectru m of cycles respon sible for alterin g earth ’s clim ate (Table 1). With lon ger periodicities, th e scale of en viron m en tal im pact becom es m ore extrem e. Th u s, wh ile season al variation s in su n ligh t, rain fall, an d tem peratu re affect m an y organ ism s, en tire lan dscapes, in clu din g th e distribu tion s of water, trees, an d an im al popu lation s, h ave been recon stitu ted by arid-wet oscillation s tim ed with th e 19-kyr an d 23-kyr precession al variation s in E arth ’s orbit. Th e glacial cycles, wh ich occu r every 100 kyr, correspon din g with th e period of orbital eccen tricity, h ave im posed an even h arsh er oscillation on th e en viron m en ts of n orth ern con tin en ts. Between th e season al an d th e orbital scales, m an y oth er cycles of diverse wavelen gth h ave been docu m en ted based on localized records of fossil pollen , sedim en tation , an d tree growth rin gs. Th is spectru m of cycles represen ts a set of sim u ltan eou s wavelen gth s th at cou ple an d decou ple, m esh an d u n m esh , som etim es a m p lifyin g a n d som etim es dam pen in g th e clim atic effects of on e an oth er. Th e cycles in teract in ways th at can lead to n on lin ear or th resh old-type ch an ge. Su ch ch an ge can in clu de decade-scale oscillation s between glacial an d warm con dition s (e.g., th e ‘‘flickerin g switch ’’ of clim ate alteration recorded in th e Green lan d ice cores 48 ) between lon g periods of ARTI CLES TABLE 1. Part of the Spectrum of Quaternary Environmental Oscillation, Indicating Cyclicity (Years) and the Associated Cause or Effect of Variation (Variable)39 Years Variable 1 2–7 22 70 125 600 (1,150, 1,700) 7,000 Annual seasons El Niño Solar variations Warming/cooling Warming/cooling Warming/cooling 8,800 (9,200) 10,500 23,000 41,000 100,000 Heinrich events (iceberg discharge) Vegetation change Ocean carbonate fluctation Orbital variation (precession) Orbital variation (obliquity) Orbital variation (eccentricity) stability. Th ey can also in clu de cen tu rylon g sh ifts between cold steppe an d warm , forested con dition s followin g tim es of h abitat stability, as recorded in Qu atern ary pollen records of E u rope an d North Am erica.49–51 Wh ile th e idea of ‘‘cycles’’ seem s to im ply predictability, paleoclim atologists n ow recogn ize th at th e cycles com e in an d ou t of ph ase, an d th at th e tem poral relation sh ip between an y two cycles (e.g., tropical m on soon an d glacial cycles) is con tin u ou sly sh iftin g.52,53 Th is com plex in teraction m ean s th at th e im pact of an y given en viron m en tal variable, su ch as solar radiation du rin g th e 100-kyr-lon g cycle of orbital eccen tricity, m ay ch an ge con siderably between on e cycle an d th e n ext. Depen din g on local topograph y, vegetation , an d latitu de, global clim ate factor s wer e, m or eover, r egister ed in differen t ways across th e globe, resu ltin g in geograph ic variation in th e tim in g an d degree of en viron m en tal alteration . For Pleistocen e Africa, su rface water an d vegetation m ay h ave been ‘‘ou t of syn ch .’’Th is is su ggested by th e stran ge discrepan cy between lake levels, wh ich vary on th e precession al cycle, an d con tin en tal du st records, wh ich correlate with th e lon ger eccen tricity cycle.54 Volcan ism an d tecton ism h ad stron g loca l a n d r egion a l in flu en ces, a n d added greatly to th e idiosyn cracy of Pliocen e an d Pleistocen e en viron m en tal ch an ge. E ru ption s an d earth qu akes were respon sible for large, repetitive, an d n early in stan tan eou s ch an ges in h ydrology an d soil ch em istry in som e areas in h abited by h om in ids, su ch as th e Tu r ka n a a n d Olor gesa ilie b a sin s.55–57 By its effect on atm osph eric circu lation pattern s, large-scale tecton ism like th at leadin g to th e H im alayan an d Tibetan Plateau u plift also en h an ced clim ate variability, for exam ple by stren gth en in g th e m on soon al variation s in Africa.58 BUFFERING OF ENVIRONMENTAL DISPARITY Th e adaptive con dition s faced by Pleistocen e organ ism s were th erefore n ot especially con sisten t. Th e largest rem odelin gs of lan dscapes were expressed over lon g span s relative to in dividu al lifetim es. E pisodes of exten sive ch an ge in vegetation , water, an d oth er resou rces occu rred between periods of relative stability. Rare, m ajor sh ifts som etim es occu rred very rapidly. In gen eral, organ ism s can evolve two ways of respon din g to h abitat ch an ge. Th e first is by m obility or dispersal, ways of trackin g a preferred h abitat or key resou rce. Most organ ism s are qu ite specific to h abitat or are depen den t on a particu lar set of life con dition s, an d m u st keep con tact with key resou rces du rin g periods of large h abitat ch an ge. By dispersal an d m obility, th ese organ ism s h ave, in a sen se, stabilized im portan t su rvival factors an d ou tcom es of n atu ral selection over tim e, wh ich is essen tial in cases of in terspecific depen den ce or coevolu tion . Th e secon d approach is to broaden th e ran ge of con dition s in wh ich an organ ism can live. Adaptive flexibility can be ach ieved in several ways. On e is by gen etic polym orph ism , in wh ich altern ative alleles con fer differen t ph en otypic an d adaptive ben efits. As en vir on m en ta l ch a n ge occu r s, th e fr equ en cy of alleles also ch an ges as a resu lt of selection at a particu lar tim e an d place. Th e advan tage of polym orph ism is th at a sh ift in allele frequ en cy en ables a popu lation of organ ism s to disperse over con trastin g settin gs or Evolutionary Anthropology 85 ARTI CLES TABLE 2. Comparison Between Extinct Large Herbivores of Southern Kenya and Closely Related Surviving Taxa Last Record in Southern Extinct Taxon Kenya (ka) Equus oldowayensis 780 Theropithecus oswaldi 780–601 Hippopotamus gorgops 601–493 Elephas recki 662–601 Metridiochoerus spp. (M. hopwoodi, compactus & cf. modestus) 990–780 Related Means of Ecological Evidence of Specialization Extant Taxon Flexibility Absolute and relative muzzle breadth larger and beyond the range of modern E. grevyi very broad and straight incisal cropping mechanism; large cheek tooth volume; more hypsodont than extant zebras. Enormous body size (males .58kg); large molars and molarized premolars with complex enamel shearing surfaces. Progressive trend to greater body size, molar occlusal area and enamel complexity near last appearance. Dental microwear indicative of less varied diet than earlier Theropithecus. Hands and feet with extreme terrestrial adaptations. Larger body and craniodental complex than in modern Hippopotamus; elevated orbits suggest specialized aquatic adaptation Equus grevyi Flexible social organization responsive to foraging conditions; capable of browsing. Papio anubis Omnivorous diet; body size ,35 kg; arboreal and terrestrial forager; variable social group size. Hippopotamus Long-distance movement between water bodies; highly flexible social amphibius grouping and territoriality. Specialized features related to Loxodonta Less specialized molars; able to dental shearing of coarse vegetabrowse and graze; long-distance africana tion (increased number of enamel tracking of resources; inhabits plates/tooth, shortened mandesert to forested habitats. dibular corpus and symphysis compared with earlier E. recki); body size and dental occlusal area much larger than in modern elephants; known distribution in savanna. Extremely hypsodont M3s; M1–3 Phacochoerus Widespread inhabitant of very dry to occlusal area up to 2.33 that moist savanna; not highly depenaethiopicus dent on water; a grazer but eats of Phacochoerus; bodies of M. fruits, shrubs, bulbs, tubers. compactus and M. hopwoodi very massive (largest metridiochoeres). accom m odate to a specific array of flu ctu atin g con dition s over tim e. Adaptive flexibility can also be attain ed by widen in g th e gen etic n orm of reaction . In th is case, exposu re to diverse en viron m en ts du rin g growth an d m atu ration elicits a ran ge of developm en tal respon ses an d ph en otypes from a sin gle gen otype. Th e effect, kn own a s p h en otyp ic p la sticity, is widely kn own , an d is a respon se to som e ran ge of en viron m en tal variation n orm ally experien ced by a popu lation .59,60 A th ird aven u e of flexibility is afforded by com plex stru ctu res or beh aviors th at are design ed to respon d to n ovel an d u n predictable adaptive settin gs. Variability selection is th e term applied specifically to th e evolu tion of th ese a d a p tive m ech a n ism s, wh ich em er ge a s a p op u la tion con fr on ts h igh ly va r ia b le en vir on m en ts over m an y gen eration s.39,40 Variability selection th u s h igh ligh ts th e effect of in con sisten t selection regim es over lon g tim e span s. Va r ia b ility selection a d a p ta tion s em erge as gen etic variation s assist an organ ism ’s soph isticated in take of an d respon siven ess to en viron m en tal data. Th e stru ctu res an d beh avioral m ech an ism s bu ilt by variability selection h ave strikin g u n iform ity with in a species, yet are able to m ediate secon dary ph en otypic traits th at vary as a species en cou n ters n ew extern al data (e.g., ecological, social). E xam ples in clu de: ● a locom otor system (prim ary, u n iversal ch aracteristic) design ed to allow a wide repertoire of m ovem en t (secon dary ph en otypic flexibility). ● a particu lar den tal stru ctu re or foragin g strategy th at en h an ces a sh ift to n ewly available food types. ● a large brain or specific n eu rological stru ctu re th at is effective in processin g extern al data an d gen eratin g com plex cogn itive respon ses. ● species-specific social beh aviors from wh ich a broad ran ge of m atin g 86 Evolutionary Anthropology stru ctu res or dem ograph ic grou pin gs m ay be m an ifested. Th e prim ary ch aracteristics are, accordin g to th e variability-selection h ypothesis, adaptive with respect to environm ental instability (the inconsistency of selective conditions) rath er th an to an y sin gle relatively sh ort-lived stable h abitat or an y series of su ch h abitats exp er ien ced over tim e. Va r ia b ilityselection adaptation s resu lt in flexible, n ovel respon ses to su rrou n din gs an d diversity in th e adaptive repertoire. Large Mammals of Southern Kenya Th e con cept of variability selection origin ally grew ou t of a stu dy of Pleistocen e large m am m als in th e sou th ern Ken ya rift. Rich fossil assem blages h ave been recovered from th e Ken yan sites of Olorgesailie an d Lain yam ok, dated ca. 992 to 350 ka.61,62 Th ese assem blages sh ow th at at least five large-bodied h erbivore lin eages, wh ich h ad on ce dom in ated th e m id-Pleistocen e fau n a of sou th ern Ken ya, becam e extin ct between 780 an d 350 ka (Table 2). Th ese h erbivores were su rvived by closely related species (sister taxa) th at, prior to 350 ka, were poorly represen ted in th e fau n a of th e sou th ern Ken ya rift. Th e extin ct form s—th e zebra Equ u s oldow ayen sis, th e m on key Theropithecu s osw aldi, th e eleph an t Elephas recki, th e h ippo Hippopotam u s gorgops, an d th e h u ge m etridioch oere pigs—h ad several traits in com m on . Th ey all h ad very large bodies com pared to earlier m em bers of th eir clades, an d th ey possessed stru ctu ral traits su ch as m assive cran ioden tal com plexes th at in dicate h igh ly specialized grazin g. Th e extan t taxa, by con trast, all exh ibit som e degree of dietary or h abitat diversity, su ch as an ability to switch to oth er foods wh en favored on es are absen t. Overall body size is sm aller, as is th e ch ewin g apparatu s, su ggestin g th at th ese taxa are less depen den t on large qu an tities of low-qu ality forage. By observation , we also kn ow th at th ese su rvivin g species are flexible in th eir pattern s of social grou pin g, an d th at grou p size varies widely in relation to available resou rces. E ven th e grazer Equ u s grevyi ch an ges its dem ograph ic an d social grou pin g pattern s, ARTI CLES an d can sh ift to eatin g browse in respon se to en viron m en tal ch an ge.63 By 350 ka, th e su ite of extin ct h erbivores h ad been replaced in sou th ern Ken ya by th e extan t taxa. 62 Wh ile stu dyin g th e possible cau ses of th is ch an ge, it becam e clear th at sou th ern Ken ya experien ced a wide ran ge of en viron m en tal flu ctu ation s du rin g th is period. At Olorgesailie, m ajor rem odelin g of th e lan dscape, in clu din g dram atic ch an ges in basin h ydrology, occu rred, on average, of every 3.6 to 65 kyr, depen din g on th e particu lar stratigraph ic in terval.57,64,65 Tu rn over in th e m am m als coin cided with extrem e m idPleistocen e flu ctu ation . Based on th is correlation an d a com p a r ison of a d a p tive ch a r a cter istics (Table 2), th e livin g taxa appear to h ave persisted becau se th ey possessed traits th at equ ipped th em to deal with th e variability of m id-Pleistocen e h abitats. Th e extin ct form s, by con trast, all h ad traits favored in dry, grass-rich en viron m en ts. An im als well equ ipped to deal with open savan n a settin gs, th e h a b ita t d eem ed cr u cia l in con ven tion al scen arios of h om in id evolu tion , m et th eir dem ise, wh ereas lin eages equ ipped with th e m ean s of versatile respon se con tin u ed. It is tem ptin g to ascribe th is process of adaptive evolu tion an d lin eage sortin g to th e large ran ge of h abitat flu ctu ation eviden t du rin g th is tim e span . Th u s th e h ypoth esis of variability selection can be stated as follows. (1) Over a len gth y span of tim e, organ ism s en cou n ter periodic rem odelin g of th e lan dscape, resou rces, an d adaptive settin g. Som e popu lation s ten d to m ove an d th u s m ain tain h abitat-specific adaptation s. (2) In oth ers cases, differen t in dividu als of th e sam e popu lation or lin eage m eet in creasin gly diverse con dition s of reprodu ctive su ccess an d su rvival over tim e. In th e latter popu lation s, th e episodic sh iftin g of selective con dition s favors an y available gen es an d related ph en otypes th at h elp to bu ild adaptive versatility. (3) As th e degree or com plexity of en vir on m en ta l r em od elin g in creases, lin eages with h abitat-specific adaptation s m ay becom e extin ct at th e expen se of popu lation s with adaptive versatility. Th e exam ple of m am m als in th e sou th ern Ken ya rift serves two pu rposes. First, it illu strates th at variabil- ity selection as an evolu tion ary process depen ds on its application to a ran ge of organ ism s, even th ou gh con sisten t selection over tim e, h abitatspecific in its effect, applies to m ost or ga n ism s. In d eed , th e va r ia b ilityselection h ypoth esis does n ot m ean th a t h a b ita t sp ecia lists u su a lly b ecom e extin ct, as eviden ced by th e m an y specialized form s livin g today. Secon d, th e exam ple illu strates th e lin eage-level sortin g con veyed by th e th ird part of th e h ypoth esis. Th e first two parts of th e h ypoth esis, h owever, lead to an oth er qu estion : H ow does variability selection operate at th e Men delian level as a way of evolvin g adaptive versatility? THE PROCESS OF VARIABILITY SELECTION A Simple Model A con ceptu al m odel of variability selection in a Men delian popu lation is presen ted in Figu re 2. It illu strates th at in con sisten cy in fitn ess con dition s m ay resu lt in a powerfu l selective process. Th at process requ ires th e presen ce of certain kin ds of gen etic variation s. It fu rth er assu m es th at th ere is m ore th an on e gen etic or developm en tal path toward su rvival an d reprodu ctive su ccess in a given en viron m en t. Th e first type of gen etic variation does best (h igh er relative fitn ess) in certain specific h abitats. Th at is, it en h an ces fu n ction s th at are well m atch ed to particu lar su rrou n din gs. Th e secon d type of gen etic variation en ables su rvival an d reprodu ction in th e in itial h abitat in wh ich it appeared an d also h appen s to do so wh en th e su rrou n din gs ch an ge. With ou t delvin g in to th e in tricate possibilities, th e on ly way we can tell wh ich gen etic varian ts will su cceed is by ru n n in g th e system for a tim e. In direction al selection , th e lon g-term resu lt is con sisten t with th e sh orter span s of selection . Th e first type of gen etic option will ou tcom pete th e secon d. Com plex system s of alleles an d fu n ction s m ay be bu ilt u p du e to th e con sisten cy of selection over a lon g tim e. Un der variability selection , by con trast, th e secon d type of gen etic varian t u ltim ately su cceeds. Th is resu lt is n ot predictable from th e tem porary a d va n ta ges con fer r ed d u r in g sh or t span s of selection in relatively stable Evolutionary Anthropology 87 ARTI CLES Figure 2. Variability selection in a Mendelian population (based on Potts40). Pleistocene environmental oscillation (Climate) causes fluctuation in vegetation and surface water (Landscape). At t1, a modest-sized lake is fed by a meandering fluvial system in open woodland. After a series of wet phases, when tree cover and surface water are abundant, and dry phases, when grassland and low lake levels prevail, surface water disappears from the area (t7). Shifts in important parameters of natural selection such as food distribution influence the genetic options of a population. The model assumes the existence of multiple genetic, developmental, and phenotypic options. Over time, three alleles, not necessarily at the same locus, play critical roles: M affects a structure or function that offers competitive advantage in moist, highly vegetated settings. Allele A governs a behavior that yields higher fitness in dry, open habitats. And C, though at greater metabolic cost than the other alleles impose, contributes to the organism’s versatility in response to resource changes by, for example, increasing digestive flexibility or providing larger memory regarding alternative food sources. The size of the shapes under Gene Pool depicts the relative fitness of these three alleles in specific time intervals. Significant adaptive change is usually thought to arise by consistent selection from one generation to another (e.g., t1 to t2). M and A have periods of high relative fitness within certain environments. Environmental oscillation and related inconsistency in selective conditions, however, ultimately favor versatility (fixation of allele C) at the expense of habitat-specific responses (elimination of A and M). By t5 to t7, the relative advantage of allele C is apparent but cannot be extrapolated from the short-term fitness results of any single interval earlier in time. Over several hundred thousand years of oscillation, a wide suite of alleles affecting multiple aspects of an organism’s diet, cognition, sociality, and other functions may become dominant because they enhance environmental data processing and novel problem solving. settin gs. E ven tu ally, com bin ation s of alleles m ay be favored th at bu ild u p specific organ s an d beh aviors th at en h an ce versatile or n ovel respon ses. Th is lon g-term process can n ot be dedu ced from th e com petition th at takes place in a specific or direction ally ch an gin g en viron m en t. As th e m odel sh ows, alleles favored by variability selection m u st avoid extin ction in each specific settin g th ey en cou n ter. Th at certain alleles m ay n ot, at first, be associated with th e h igh est reprodu ctive su ccess an d su rvivorsh ip n eed n ot both er u s as lon g as th ey perm it m odest su rvivorsh ip in a given en viron m en t an d as lon g as episodic ch an ge in th e su rrou n din gs u ltim ately depresses th e fitn ess of h abitatspecific alleles th at on ce h ad a greater advan tage. In fact, th e m ain ten an ce of diverse ph en otypes in a popu lation , som e with lower fitn ess th an oth ers (e.g., deviation s from optim al egg size in vertebrates 66 ), h as been docu m en ted in a variety of organ ism s.67 Th e persisten ce of lower-fitn ess in dividu als or popu lation s, m oreover, is cen tral to th e idea of th e fitn ess lan dscape, wh ere som e in dividu als or popu lation s lie on lower adaptive peaks th an oth ers do, yet rem ain viable u n der th e con dition s at h an d (see Box). Th e adaptive attribu tes evolved by variability selection are h eritable, yet th e su rvival advan tages th ey con fer are eviden t on ly after a period of m arked in con sisten cy in relative fitn ess. Tradition al ideas abou t n atu ral selection em ph asize th e close m atch th at evolves between an organ ism an d its specific en viron m en t. Most organ ism s are in deed h abitat specialists, each tied to a specific set of adaptive con dition s. Th is close m atch in g is wh at 88 Evolutionary Anthropology ARTI CLES Variability Selection and Wright’s Adaptive Landscape Variability selection has a basis in existing evolutionary theory, yet can be considered a unique process. These two points are illustrated by shifting balance theory, developed originally by Sewall Wright.91,92 Wright envisioned species as being divided into small, semi-isolated populations. Adaptive change resulted from the balance among random drift and local selection (these, together, tended to diversify the available systems of interacting genes), migration, and ‘‘interdemic selection’’—the favoring of certain gene systems over other varieties spread through different populations. Shifts in the strength of these factors, according to Wright, caused some populations to go extinct but enabled others to traverse a valley of lower fitness and ascend an adaptive peak of high selective value, as represented in his famous image of the adaptive landscape (see Fig.). Wright’s theory differed substantially from those of others. It particularly differed from that of Fisher, whose treatment of evolution assumed large, panmictic populations; a constant supply of favorable mutations; the slow, constant shifting of environmental conditions; and the power of consistent, directional selection to create adaptive change.93,94 Shifting balance theory anticipates many of the critical elements of variability selection. An irregularly changing environment is one of several factors Wright considered in his explanation of how adaptive complexity evolves. In contrast with Fisher and other contemporaries, Wright believed that adaptive complexity cannot arise under steady selection pressure. Environmental change is one way of varying selection. It creates shifts in the peaks and valleys of the fitness surface, and thus is a way of experimenting with the adaptive success of diverse gene frequencies and complexes of interacting alleles, which are represented on the slopes of peaks and in the valleys. The interaction between individual selection and population structure is, moreover, vital to shifting balance theory in that populations or lineages with different histories develop different adaptive properties. Extinction is likely in a population or species that occupies a narrow, even if high, adaptive peak, which Wright referred to as ‘‘the victim of extreme specialization to conditions which have ceased.’’91 The central idea of variability selection— that disparity in selective conditions ultimately favors key gene complexes of lower original fitness than those that flourish in specific environments—is analogous to the passage of a population or lineage through adaptive troughs on the way to higher adaptive peaks (see legend Fig. 2). Thus, key ingredients of variability selection, among them variable adaptive settings, selective effects at the Mendelian level, persistence of novel gene complexes through periods of lower fitness, and expression of the results as differential survival of lineages, are all consistent with Wright’s intricate theory of evolution. The idea of variability selection differs in several respects from Wright’s focus, however, and draws attention to factors that could be incorporated into contemporary theory. First, although he strongly considered it, Wright did not deem environmental change to be essential for creating inconsistency in natural selection. The original shiftingbalance theory required only that populations be divided and that new complexes of interacting genes be developed in order for complex adaptive evolution, through interdemic selection, to take place. Wright placed strong emphasis on random shifts in gene frequencies as a way of creating these new gene interaction systems on which selection would act. In no manner does the idea of variability selection dispute Wright’s mechanism or the variety of ways by which diversity in genic systems may develop. But it does give central importance to periodic remodeling of environments and adaptive settings. Without this remodeling, it would be impossible for selection to build gene complexes that enable an organism to respond to complex environmental disparities and that consequently permit its lin- A: Wright’s classic depiction of an adaptive landscape, on which 1 signifies a gene combination of superior fitness (an adaptive peak) and 2 signifies lower fitness (an adaptive valley). B: A cross-section of the adaptive landscape and Wright’s indication (arrow) of the central problem of adaptive evolution, which is how a population can traverse an adaptive valley between a relatively low and a higher adaptive peak. eage to accommodate over the long term. A second, related difference concerns the rate and degree of environmental alteration. The large-scale, episodic change vital to variability selection contrasts with the smoother movement of adaptive peaks and valleys and the shuffling of populations typically envisioned on Wright’s fitness surface. According to variability selection, habitat instability leads to irregular, dramatic, and often rapid reconfiguring of the adaptive surface. Relatively long-lived peaks become valleys, and vice versa. Most treatments of shifting balance theory, by contrast, deal with seasonal or annual fluctuations and how populations adjust by modifying the frequencies of alleles. The temporal perspective and degree of alteration posited by the variability-selection hypothesis is thus far untested in the framework of shifting balance theory. (Continued) Evolutionary Anthropology 89 ARTI CLES Variability Selection and Wright’s Adaptive Landscape (continued) Third, as just noted, most applications of shifting balance theory since the 1950s have considered selective inconsistency only in terms of polymorphic loci. That is, shifts in environment and selection lead to polymorphism and change in allelic frequencies over time and place. In variability selection, by contrast, selective disparities build up gene complexes that enable an m ost biologists m ean by adaptation , an d it im plies th e in flu en ce of direction al selection . Th e variability-selection m odel, by con trast, stresses th at a lin eage m ay experien ce factors over tim e th at disru pt its close tie to an y specific en viron m en t. Th e resu lt is an even tu al decou plin g of th e organ ism from an y on e en viron m en tal state. In sh ort, variability selection is a tim e-in tegrated selective process by wh ich adaptive m ech an ism s em erge th at bu ffer episodic ch an ge in an en viron m en tal settin g. It is an aspect of n atu ral selection , fu lfillin g th e th ree n ecessary an d su fficien t con dition s n oted earlier, bu t with ou t th e assu m ption of lon g-term con sisten cy in selective resu lts. Th e ph en otypic resu lts of th e process can n ot be extrapolated from an y sin gle gen eration , or several gen eration s, of Darwin ian selection . Its effect occu rs in th e con text of lon gterm , recu rren t rem odelin g of en viron m en ts su ch th at in dividu als of a lin eage (a popu lation or su bdivided set of popu lation s) experien ce m axim ally differen t adaptive con dition s at widely differen t tim es. As a resu lt, certain gen etic com b in a tion s a n d d evelop m en tal pattern s are favored th at bu ild m ech an ism s design ed for com plex respon se to n ovel or u n predictable en viron m en tal in form ation . Th e prim ary ph en otypes favored by variability selection are in tricate, can alized adaptive system s th at com e to typify th e lin eage. Th e secon dary ph en otypic ou tpu t of th ese system s is beh avioral versatility an d diversity. Potential Difficulties of the Idea Th e m odel (Fig. 2) an d basic idea of variability selection h ave arou sed cer- organism to respond to novel adaptive situations and the organism’s population to endure over time without necessarily further altering its genetic constitution. The only way to envision this using Wright’s adaptive landscape is as a time series in which the disturbance of peaks and valleys begins to favor any system of interacting genes that, in a sense, stabilizes or flattens tain dou bts an d qu estion s, th e m ost com m on of wh ich are con sidered h ere: 1. Isn ’t variability selection already accou n ted for by ‘‘flu ctu atin g selection ,’’ a term u sed to explain th e developm en t of gen etic polym orph ism an d ph en otypic plasticity? As n oted earlier, gen etic polym orph ism an d ph en otypic plasticity are ways of attain in g adaptive flexibility. Math em atical m odels an d sim u lation s h ave sh own h ow th ese ph en om en a m ay arise wh ere th e en viron m en t var- . . . by contrast, the variability-selection model stresses that a lineage may experience factors over time that disrupt its close tie to any specific environment. The result is an eventual decoupling of the organism from any one environmental state. ies between certain optim al valu es, each favorin g an altern ative ph en otype (flu ctu atin g selection ).60,68,69 In p h en otyp ic p la sticity, for exa m p le, ‘‘som e cu e m u st sign al th e state of th e en viron m en t so th at a direction al respon se can be in du ced,’’60 en ablin g th e developm en t of on e ph en otype over an oth er. In addition , em pirical stu dies of flu ctu atin g selection in livin g organ - its particular area of the fitness surface. In other words, certain genic systems eventually buffer a population from being cast into a deep adaptive valley, even while other gene complexes, those tuned to more specific adaptive settings, remain susceptible to large peak-and-valley disturbances related to environmental change. ism s h ave, by n ecessity, focu sed on sh ort tim e fram es, su ch as an n u al variation s in h arsh season s.66,69,70 In gen eral, it is believed th at th e effects of en vir on m en ta l flu ctu a tion m u st b e scaled to an organ ism ’s expected lifetim e an d developm en tal sch edu le if an a d a p tive in flu en ce is to b e m a n ifested.71 Th e variability-selection h ypoth esis, by con trast, posits th at a m ore com plicated, lon ger-term en viron m en tal h istory m ay also affect th e adaptive properties of a lin eage of organ ism s. Su ch selection m ay occu r wh ere n o sin gle com bin ation of factors (for exam ple, food abu n dan ce, com petitors, predators, an d parasites) th at an organ ism experien ces over several con secu tive gen eration s is replicated in distan t gen eration s. As wider disparities are m an ifested, m ech an ism s m ay evolve th at foster adaptive versatility an d th at are design ed to en able th e organ ism to cope with large, u n predictable en viron m en tal variability. Th ese adaptation s, in oth er words, are n ot based on gen etic polym orph ism or in creased reaction n orm s. In deed, th e m ean s of adaptive flexibility evolved by variability selection becom e decou pled from an y specific set or range of environm ents. This decoupling raises the possibility of novel response to adaptive problem s. Un like flu ctu atin g selection , variability selection posits th at lon g periodicities of en viron m en tal ch an ge are essen tia l in evolvin g th ese d istin ctive adaptive m ech an ism s. It assu m es th at disparity in selective con dition s becom es m ore severe over exten ded periods, an d th at th is in creasin g disson an ce can h ave a stron g effect on th e adaptive properties of a lin eage. Th e 90 Evolutionary Anthropology variability-selection process th u s accords an im portan t biological role to en viron m en tal flu ctu ation th at is m easu rable in th e geologic record. 2. Isn ’t season ality or en viron m en tal variation within an organism ’s lifetim e sufficient to account for the type of versatility ascribed to variability selection? Paleoen viron m en tal records coverin g th e past several m illion years sh ow con vin cin gly th at ch an ge in h abitats an d adaptive con dition s becom es m ore severe over exten ded periods. Th at is, or b ita l-sca le va r ia tion s a lter la n d scapes an d resou rces to a far greater exten t th an do season al or an n u al variation s. Th e latter m ay in volve sign ifica n t sh ifts in tem p er a tu r e a n d m oistu re, bu t do n ot cau se th e overall revision of region al an d con tin en tal h yd r ology, vegeta tion , a n d a n im a l com m u n ities eviden t over span s of 10 3 to 10 5 years. Lan dscapes u n dergo exten sive r em od elin g r a th er sp or a d ically. In dividu al soils an d particu lar sets of ch an n el, floodplain , or lake sedim en ts, for exam ple, u su ally en com pass a few h u n dred to several ten s of th ou san ds of years between th e periods of sign ifican t ch an ge in deposition al an d ph ysical en viron m en t.72,73 3. Isn ’t variability selection m erely th e su m of su ccessive periods of direction al selection ? It is possible to m isin terpret th e m odel in Figu re 2 in term s of su ccessive periods of direction al selection , a sequ en ce th at wou ld resu lt in th e com pilation of a poten tially large set of h abitat-specific adaptation s. In th at case, particu lar h abitat-orien ted beh aviors an d traits wou ld be elicited by particu lar en viron m en tal states. A differen t kin d of adaptive system is su ggested to resu lt from variability selection , h owever. Accor d in g to th e variability-selection h ypoth esis, large en vir on m en ta l in con sisten cies over tim e favor specific beh avioral or an atom ical com plexes th at are especially su ccessfu l in n ovel settin gs an d can h elp to bu ffer th e in con sisten cies. In oth er words, th ese traits are bu ilt u p becau se th e u n derlyin g alleles con tribu te especially well to an organ ism ’s su r viva l a n d r ep r od u ctive su ccess wh en ever la n d sca p es u n d er go ep isodes of ch an ge. Th e resu lt is a decou plin g of th e organ ism from an y on e h abitat or set of particu lar en viron m en tal states. Variability-selection ad- ARTI CLES aptations thus differ from the adding up of habitat-specific adaptations over successive periods of directional selection. 4. H ow is it possible for th e gen etic variation s su pposedly favored by variability selection to persist with ou t bein g swam ped by m ore im m ediate selection pressu res? As n oted earlier, th ere is com pellin g eviden ce in m an y differen t organ ism s th at altern ative ph en otypes, in clu din g on es of lower fitn ess, can be m ain tain ed in a popu lation over tim e.67 Lower-fitn ess in dividu als m ay redu ce com petition with con specifics, for exam ple, by foragin g for less favored foods or occu pyin g th e periph ery of th e h om e ran ge. Th ey are th u s able to m ain tain viable offsprin g, wh ich m ay h ave som e advan tage wh en en viron m en tal con dition s ch an ge. Th e effects of im m ediate com petition an d selection pressu re m ay th erefore be m oderated by h ow a popu lation is spatially su bdivided an d distribu ted. Wh eth er th e process of variability selection presen ted h ere is feasible or n ot still requ ires m ath em atical m odelin g an d testin g by sim u lation . At least th ree types of variables n eed to be in clu ded in a qu an titative m odel of th e process: (a) Th e tim e scale of variability: Gen etic ch an ge depen ds on con tin u ity from on e gen eration to th e n ext. Yet, accordin g to th e variability-selection h ypoth esis, adaptive ch an ge m ay arise from large disparities in selective con dition s expressed over span s far lon ger th an a gen eration or an organ ism ’s lifetim e. Th is apparen t con flict m ay be resolved by th e existen ce of a graded spectru m of en viron m en tal variability, from season al to orbital tim e scales (Table 1). Th is gradation provides a bridge between gen eration -scale gen etic ch an ge an d th e len gth ier tim e fram e of m assive en viron m en tal rem odelin g. Variability selection can n ot work if its selective effects are felt on ly on ce every 10,000 years. Th e process depen ds on th e spectru m of en viron m en tal variability. A gen etic variation favored at on e scale of en viron m en tal variability m ay yield a con com itan t advan tage at sligh tly lon ger an d m ore extr em e sca les of va r ia b ility. With lon ger tim e fram es, th e gen e pool en cou n ters a gradation al widen in g of en viron m en tal variability, pu n ctu ated by episodes of m ore extrem e ch an ge an d in stability. Accordin g to variability selection , cer ta in gen etic tr a its tu n ed to th e tem po an d degree of m ajor en viron m en tal revision are even tu ally favored over th ose tu n ed to particu lar en viron m en ts or h abitat-specific selection pressu res. Th e rate of m ajor en viron m en tal revision , gradin g of th e spectru m of variability, an d gen eration tim e all are likely to be in volved. Th e feasibility of variability selection m ay depen d on a n arrow set of in ter a ction s a m on g th ese th r ee ‘‘tem po variables.’’ (b) Types of gen etic variation . Wh at kin d of gen etic variation s an d m ech an ism s m igh t be in volved in variability selection ? Th e proposed process h as several requ irem en ts th at h in t at an an swer to th is im portan t qu estion . In variability selection , th e favored a lleles foster p a r ticu la r str u ctu r es an d beh aviors th at en h an ce an organ ism ’s respon siven ess to en viron m en tal ch an ge, in clu din g its capacity for com plex in form ation processin g. Th e con tribu tin g gen es wou ld h ave to form an in tegrated com plex. Mech an ism s su ch as lin kage, wh ich en h an ce th e in tern al coh esion of th ese gen es, probably n eed to be in volved. We m ay also n ote Levin s’ u sefu l term ‘‘u n bin din g’’74 to den ote th at, as variability selection proceeds, an organ ism ’s adaptive system s becom e less tied to on e an oth er an d to specific su rrou n din gs. Accordin g to Levin s, ‘‘p h yletic lin es wh ich h a ve p a ssed th rou gh m an y differen t kin ds of en viron m en ts on a geological tim e scale will be con stan tly bin din g an d u n bin din g [gen etic, developm en tal, an d ph en otyp ic] va r ia b les, will h a ve m or e loosely b ou n d xi [p h en otyp ic va r iables] an d m ore degrees of freedom , an d will th erefore be m ore able to u n dergo m ajor restru ctu rin g.’’In developm en tal biology, dissociation sim ilarly refers to th e u n lin kin g of developm en tal system s. Th is m ean s th at as on e adaptive m ech an ism is altered it does n ot force con com itan t, possibly d eleter iou s, a lter a tion s in oth er s. 75 Th is type of ch an ge in developm en t seem s to be critical to th e evolu tion of variability-selection adaptation s. Coordin ation of gen e expression is th e obviou s keyston e. H om eobox gen es appear too h igh ly con served across taxon om ic grou ps to h ave been in volved directly. Gen es th at con trol de- Evolutionary Anthropology 91 ARTI CLES velopm en t are, h owever, in tegral to an y evolu tion ary process as com plex as variability selection . It rem ain s to be seen exactly h ow su ch effects as u n bin din g, dissociation , an d th e in tegration of fu n ction ally lin ked gen e arrays m igh t arise. At a m in im u m , regu latory gen etic ch an ge, with a stron g effect on th e tim in g of developm en tal processes, is fu n dam en tal to an y lin eage th at h as been su bject to variability selection . (c) Species popu lation stru ctu re. Differen t popu lation s of a species m ay con fron t differen t ran ges an d pattern s of en viron m en tal variability over tim e an d space. Du e to sem i-isolation , th ey m ay also develop distin ctive allele com bin ation s. Gen e flow, th e tran sfer between popu lation s of th ese allele com bin ation s, m ay allow m ore effective respon ses to large an d u n expected sh ifts in su rrou n din gs. Th e su bdivision of a lin eage in to popu lation s also m u ltiplies th e n u m ber of experim en ts in volved in respon din g to adaptive disparity. Popu lation size an d stru ctu re th u s are critical factors in th e operation of variability selection , affectin g th e viability of in itially lower-fitn ess alleles u n til th ey even tu ally replace h abitat-specific on es an d th e differen tial su ccess of popu lation s in su rvivin g a given sh ift in adaptive con dition s. 5. Becau se organ ism s can n ot be adapted to fu tu re con dition s, h ow is it possible for variability selection to operate? Th e sim ple m odel in Figu re 2 su ggests th at som e m u tation s m ay ben efit descen den ts as adaptive con dition s diversify over tim e. Th is does n ot m ean th at an organ ism is som eh ow adapted to fu tu re con dition s; th at is im possible in n atu ral-selection th eory. In stead, in th e process of variability selection , alleles can in crease on ly th rou gh a h istory of differen tial replication du e to en viron m en tal ch an ge. As is tru e of all adaptation s, traits favored by variability selection reflect a series of past even ts. Th e variability-selection h ypoth esis sees large-scale en viron m en tal alteration over a len gth y past as bein g respon sible for in creasin g certain gen etic variation s an d, u ltim ately, for bu ildin g adaptive versatility. 6. Isn ’t th e ou tcom e of variability selection already con veyed an d accou n ted for by th e gen eralist-specialist dich otom y? Variability selection resu lts in specialization s (derived featu res) th at are adapted to en viron m en tal n ovelty. Th e fam iliar distin ction between specialists an d gen eralists, th erefore, is n ot especially u sefu l in explain in g th e process an d its resu lts. Th e term ‘‘gen eralist’’ is som etim es con fu sed with prim itive m orph ology, or a lack of derived traits. Specialization , m oreover, is often con stru ed to im ply an even tu al evolu tion a r y d ea d en d . H owever, th e specialization s favored by variability selection im prove a popu lation ’s ch a n ces of su r vivin g h a b ita t a lteration s. Variability selection resu lts in organ ism s th at are specialized in dealin g with dyn am ic in form ation abou t th eir su rrou n din gs an d in bu fferin g en vir on m en ta l d isr u p tion s. Th ese sam e specialization s m ay en cou rage th e spread in to diverse h abitats, wh ich u su ally is deem ed to be a ch aracteristic of a ‘‘gen eralist.’’ Variability selection results in specializations (derived features) that are adapted to environmental novelty. Organ ism s in a popu lation affected by variability selection fu rth erm ore, are n ot n ecessarily ‘‘r-selected,’’ alth ou gh th eir popu lation s m ay be resilien t an d good colon izers.76,77 Th is resilien ce an d colon izin g ability do n ot a r ise fr om r ep r od u cin g r a p id ly, a s r-selected ‘‘weed’’ species do. Rath er, variability-selected organ ism s are resilien t for several reason s. Th ey m ay h ave th e cogn itive an d social m ean s of alterin g diet an d foragin g strategy, as does th e baboon Papio an u bis; ways of passin g on com plex social in form ation , as does th e eleph an t Loxodon ta african a; an ability to vary th eir pattern of social grou pin g, as is seen am on g popu lation s of th e zebra Equ u s grevyi); or sim ilarly in tricate ways of respon din g to en viron m en tal in su lt an d copin g with wide h abitat variation . Th u s th e u su al dich otom ies— specialist or gen eralist, sten otope or eu rytope, r-selected or K-selected— fail to captu re th e resu lts of variability selection . For wan t of a term , I su ggest ‘‘variability-selection adaptation s’’ an d ‘‘va r ia b ility-selected or ga n ism s,’’ or sim ply versatilists, to refer to th e con sequ en ces of evolu tion in respon se to lon g-term h abitat variability. 7. Is a n ew term n ecessary? ‘‘Kin selection ’’ an d ‘‘sexu al selection ’’ are term s th at were in ven ted wh en it was realized th at selective advan tage m ay exten d beyon d an in dividu al’s own reprodu ctive su ccess or econ om ic fu n ction . With sexu al selection , as Darwin recogn ized, traits m ay be favored th at con flict with econ om ic u tility or su rvivorsh ip. With kin selection , on th e oth er h an d, th e con cept of fitn ess is exten ded beyon d th e in dividu al; advan tages to kin m ay con flict with an d ou tweigh in dividu al advan tage. Becau se closer kin sh are a larger percen tage of gen es, u n u su al traits su ch as altru ism m ay evolve as a resu lt of in clu sive rath er th an in dividu al fitn ess. Both con cepts apply to th e differen tial replication of gen es in gen e pools. Th e variability-selection con cept also does, bu t exten ds th e tim e fram e, as well as th e sequ en ce of en viron m en ts, over wh ich th e biased replication of gen es is com pu ted. As we h ave seen , a con flict m ay arise between h abitat-specific traits favored in on e gen eration an d traits favored in a lon ger sequ en ce of disparate en viron m en ts. Wh ere h abitat variability is exten sive, th e key selective effect m ay occu r over a span far lon ger th an a gen eration . In variability selection , th e tim e fram e of selection m u st be exten ded. Th is is an alogou s to exten din g th e con cept of selection beyon d in dividu al fitn ess or econ om ic u tility. In addition , like th e term s kin selection an d sexu al selection , th e ‘‘variability’’ in variability selection focu ses on th e selective agen cy. Th e term refers to variability in selective con dition s (en viron m en tal variability), n ot to selection for gen etic or ph en otypic variability per se. IMPACT ON HOMINID EVOLUTION Test Predictions Alth ou gh m ost en viron m en tal h ypoth eses of h om in id evolu tion focu s on specific kin ds of h abitat, it is com m on place to observe th at at least certain h om in ids evolved wider degrees 92 Evolutionary Anthropology of flexibility in th eir en viron m en tal adaptation s th an did earlier species. Th e qu estion is wh eth er or n ot variability selection explain s th is evolu tion ary resu lt. H ow wou ld we kn ow if en viron m en tal extrem es at th e broadest tem poral scale an d related in con sisten cies in selection affected h om in id adaptive evolu tion ? Th e variability-selection h ypoth esis h as five m ain test prediction s, wh ich h elp to assess th e evolu tion ary basis of key h om in id ch aracteristics. 1. Th e h ypoth esis requ ires eviden ce th at th e featu re u n der in vestigation is design ed to en h an ce flexibility an d respon siven ess to n ovel con dition s. 2. It fu rth er requ ires th at th e featu re evolved du rin g a period of expan din g en viron m en tal flu ctu ation . If th e degree of en viron m en tal flu ctu ation rem ain ed stable, spatial variation m ay provide a m ore com pellin g explan ation of an y in crease in adaptive flexibility. 3. It predicts th at in localities with lon g stratigraph ic records, h om in id popu lation s persisted th rou gh m u ltiple, m ajor sh ifts in en viron m en t. Tools an d fossilized bon es can be u sed to in dicate th e presen ce an d persisten ce of h om in id s. If evid en ce of h om in ids appears an d disappears in correlation with specific h abitat variation s, th e h ypoth esis wou ld be n u llified. 4. If th e ran ge of h abitat flu ctu ation rose th rou gh ou t th e evolu tion ary h istory of a clade, th e h ypoth esis predicts th at adaptive versatility also becam e elaborated over tim e. Th at is, n ew m ech an ism s of respon se to en viron m en tal in stability are added to, or even replace, older m ech an ism s over tim e. In lin e with prediction 2, th e tim in g of a d d ition or r ep la cem en t sh ou ld coin cide with risin g flu ctu ation an d wider disparity in adaptive settin gs. 5. Fin ally, with in a clade, th e h ypoth esis predicts th at lin eages with h abitatspecific adaptation s ten d to go extin ct wh ile lin eages equ ipped to deal with wider en viron m en tal toleran ces persist. Th is last prediction m igh t seem to stack th e deck in favor of variability selection . Versatility always win s ou t. Bu t, in fact, m ost species in m ost clades are h abitat-specific. In oth er clades, variability selection predicts th at th ere h ave been episodes wh en ARTI CLES m ore versatile form s were favored over h abitat-specific taxa. Effects On Hominid Evolution Several an atom ical an d beh avioral featu res in h om in ids appear to satisfy som e or all of th e test prediction s an d th u s poten tially qu alify as variabilityselection adaptation s. Th ree su ch featu res are th e locom otor stru ctu re of ea r ly a u str a lop ith ecin es, b r a in en ceph alization (an d related cogn itive soph istication ) in Pleistocen e Hom o, an d th e com plex social m ech an ism s th at em erged in certain late Pleistocen e h om in ids, particu larly H. sapien s. Th e early au stralopith ecin es m an ifested a body plan related to terrestrial bipedality an d tree clim bin g. Th ey exh ibited a persisten t an d u n preceden ted m osaic of h u m an -like an d ape-like ch aracters th at in clu ded effective pelvic an d kn ee m odification s for twolegged stridin g cou pled with powerfu l forelim bs an d join t m an eu verability in dicatin g arboreal activity. Th is m orph ological th em e, wh ich persisted for two m illion years or m ore, seem s to h ave afforded locom otor versatility an d, in particu lar, an ability to accom m odate to den sely wooded an d qu ite open h abitats.39 Moreover, th e early au stralopith ecin es appear to h ave en cou n tered a variable series of en viron m en ts. Wh ile m u ch h as recen tly been m ade of th e closed vegeta tion settin gs of ea r ly h om in ids, th e few existin g Pliocen e data poin ts actu ally su ggest a ran ge from closed to relatively open h abitats (su m m arized in Potts 39). At th e Ken yan sites of Loth agam , Kan apoi, an d Tu gen H ills, in terpretation s h ave stressed th e m osaic of open an d closed vegetation between 7 an d 4 Ma. By con trast, closed woodlan d or forest h as been in ferred from eviden ce at Aram is an d Sterkfon tein . At Laetoli, a season ally dry savan n a is in dicated by th e presen ce of calich e deposits an d oth er geologic clu es. Certain aspects of th e Laetoli fau n a also offer eviden ce of a wooded h abitat, th ou gh it is im possible to tell wh eth er woodlan d coexisted with open settin gs or prevailed at differen t tim es. Based on fossil pollen , Makapan sgat h as been in terpreted as a forest, alth ou gh pollen sam ples from differen t strata at th is site su ggest con siderable flu ctu ation . Fossil pol- len , wood, an d fau n a from th e Om o an d H adar, E th iopia, fu rth er su ggest that repeated expansion, contraction, and coalescing of grasslands and woodlands were the norm during the Pliocene. If th ese com bin ed data reflect th e con texts in wh ich early au stralopith ecin es evolved, th eir locom otor system , wh ich was th eir m ost distin ctive adaptive ch aracteristic, m ay best be con sidered a respon se to growin g variability—flu ctu ation s between open an d closed settin gs—d u r in g th e la te Miocen e an d earliest Pliocen e of Africa, rath er th an an adaptation to eith er open or closed en viron m en ts. Data from late Miocen e Africa are, u n fortu n ately, qu ite rare. H owever, th e global oxygen isotope record (see Fig. 1b) sh ows a sign ifican t rise in th e scale of en viron m en tal flu ctu ation between 6 an d 5 Ma, wh ich con tin u ed at a h igh level th rou gh th e early Pliocen e. Alth ou gh by n o m ean s con clu sive, th is su ggests th at th e an atom ical pattern u n derlyin g locom otor versatility in th e earliest h om in ids arose as h abitat variability in creased. It is likely th at a m osaic of forested, wooded, an d open h a b ita ts p er sisted th r ou gh ou t th e Pliocen e. Fu rth er tests of th e variability-selection accou n t of early bipedality, th erefore, will depen d on kn owin g precise association s between h om in id fossils an d an cien t h abitats from m u ltiple localities. If th e earliest au stralopith ecin es were associated with a n arrow rath er th an an expan din g ran ge of h abitats, th e variability-selection in terpretation wou ld be falsified (test prediction 3). By 2.5 m illion years ago, som e h om in id popu lation s were m akin g ston e tools. By th e latest Pliocen e, th is beh avior in volved lon g-distan ce tran sport of ston es as far as 10 km . It also widen ed th e ran ge of accessible foods to in clu de, for exam ple, deeply bu ried tu bers, m eat, an d m arrow, an d en tailed m ore com plex spatial tasks lin kin g food to u sefu l ston es.78 Accordin g to global en viron m en tal records (Fig. 1), after 1.7 Ma, as Hom o began to th rive in in creasin gly open h abitats, h om inid populations encountered even wider am plitudes of clim ate oscillation. Du rin g th e early Pleistocen e, h om in id toolm akers occu pied a m ore varied ran ge of h abitats, m oved rocks over lon ger an d lon ger distan ces, an d exh ibited greater respon siven ess to Evolutionary Anthropology 93 ARTI CLES variation s in lan dscapes an d basin s over tim e.79 Fu rth erm ore, in early an d m iddle Pleistocen e sequ en ces, h om in id toolm akers left beh in d am ple sign s of th eir presen ce in diverse en viron m en ts, an d apparen tly accom m odated to large en viron m en tal sh ifts recorded with in particu lar locales.57,80,81 E viden ce from in dividu al sites sh ows th at late Pleistocen e popu lation s en du red even glacial-in terglacial oscillation s.82 Alth ou gh th e details h ave yet to be assem bled, it appears th at Pleistocen e h om in ids becam e less in clin ed to track particu lar h abitats as ch an ge occu rred an d m ore capable of adju stin g to n ovel con dition s an d th e in creasin g ran ge of oscillation . Th e m ost prodigiou s pace of brain en ceph alization occu rred between 600 an d 150 ka.83,84 Global isotope records an d local records from con tin en tal sites (Fig. 1) sh ow a stepwise in crease in th e am plitu de of en viron m en tal ch an ge begin n in g 640 to 700 ka. Over th e past 700 kyr, th en , between span s of relative h abitat stability, h om in ids alm ost certainly m et with strong displacem ents, shrinkages, expansions, and divisions of resources on which they relied. With regard to en ceph alization , th e va r ia b ility-selection h yp oth esis a p pears to be su pported by th e followin g observation s: Relative expan sion of th e brain en h an ced its data processin g a n d in tegr a tin g fu n ction s, wh ich greatly im proved versatility an d respon siven ess to n ovel adaptive problem s; th e largest in crease in relative brain size began with th e on set of th e largest oscillation s kn own so far in en viron m en tal records of th e past 6 m illion years (Fig. 1); an d h om in id popu lation s are eviden t du rin g th e m iddle Pleistocen e th rou gh lon g stratigraph ic sequ en ces of large-scale en viron m en tal ch an ge. Moreover, th ere appears to h ave been a su ccession of evolved m ech an ism s th at am plified th e adaptive flexibility of certain h om in id taxa over tim e (test prediction 4). Pliocen e locom otor versatility was su cceeded in th e early Pleistocen e by an expan sion of dietary possibilities, h abitat diversity, an d distan ces of m ovem en t. Th ese m ea n s of a d a p tive flexib ility wer e h eigh ten ed as relative brain size in creased du rin g th e m iddle Pleistocen e. In still later popu lation s, n ew m ean s of beh avioral flexibility were m an i- fested, in clu din g com plex sym bolic codin g, m ore rapid an d spatially diverse tech n ological in n ovation , an d powerfu lly coordin ated social action su ch as bon e arch itectu ral feats an d lon g-distan ce tradin g. Th ese n ew possibilities represen ted an u n preceden ted degree of behavioral versatility, and were expressed after several hundred thousand years of intense habitat change. Th e ratch etin g u p of adaptive versatility, th erefore, tracked th e dram atic in crease in global en viron m en tal oscillation begin n in g abou t 6 Ma, wh ich becam e extrem e over th e past 1 m illion years (Fig. 1). In fact, with in th e past 500,000 years, th e ran ge of sin gle m ajor oscillation s in th e oxygen isotope cu rve h as exceeded th e total average ch an ge over th e past 6 m illion years. E m ergen ce of th e m ost advan ced m ean s of h u m an social an d m en tal problem solvin g du rin g th is period is con sisten t with th e variability-selection h ypoth esis. Never th eless, va r ia b ility selection can , at best, be con sidered as on ly on e of m an y evolu tion ary processes th at affected h om in ids. Popu lation s were in flu en ced by a com bin ation of oth er selective param eters, in clu din g direction al, stabilizin g, sexu al, an d kin selection . Th ere are obviou s exam ples of h a b ita t-sp ecific a d a p ta tion in ea r ly h om in ids, su ch as th e lin ear body form of tropically adapted H. ergaster.85 Th u s, direction al an d stabilizin g selection occu rred even wh ile oth er types of selection were operatin g on d iffer en t fea tu r es in th e sa m e h om in ids. Th e variability-selection h ypoth esis recogn izes th at certain species evolved featu res th at were favored in specific h abitats, su ch as th e coldadapted featu res of Nean derth als, even wh ile th ey also possessed versatile, variability-selected adaptation s su ch as en larged brain s. Wh en testin g variability selection or oth er selection h ypoth eses on e m u st, th erefore, specify particu lar featu res or trait com plexes an d th en in vestigate th eir adaptive properties, as well as th e precise per iod a n d con texts in wh ich th ey em erged. Certain featu res m ay reflect th e properties of specific types of en viron m en t; oth ers m ay reflect ch an gin g settin gs, in clu d in g extr em es m a n ifested over geologic tim e. Fin ally, sortin g of h om in id lin eages (test prediction 5) su ggests a parallel with oth er clades of large Pleistocen e m am m als (Table 2). Paran thropu s boisei, for exam ple, h ad a m assive cran ioden tal apparatu s, facial bu ttressin g, an d cran ial crestin g related to specialized ch ewin g. Wh ile it m ay h ave h ad a diverse diet, Paran thropu s presu m ably h ad to activate its su bstan tial ch ewin g m ach in ery even wh en dealin g with soft foods. Its m orph ology th u s su ggests fairly special dietary requ irem en ts. Its con tem porary, H. erectu s/ ergaster, also h ad an u n u su al pattern of cran ial growth , bu t in volvin g th e n eu rocran iu m an d brain expan sion . A big brain m ay well h ave been a better organ of flexibility th an was a big ch ewin g m ach in e; all we kn ow is th at th e larger brain m orph ology was correlated, in H. erectu s, with factors su ch as wider geograph ic ran ge, wh ich m itigated th e ch an ces of extin ction . Later in th e Pleistocen e, th e Nean derth als also h ad a restricted geograph ic ran ge an d skeletal traits su ch as sh ort distal lim b segm en ts th at in dicate th eir lin k to glacial tu n dra an d tem perate (season ally cold) h abitats of western E u rasia. Th is con trasted with th e diverse h abitat of con tem poran eou s h om in ids as th ey spread to n ew areas. In both of th ese com parison s, th e pattern of lin eage extin ction an d persisten ce con form s to th e evolu tion ary respon se to wide h abitat variability predicted by th e variability-selection h ypoth esis. SOME IMPLICATIONS OF VARIABILITY SELECTION Adaptation to Environment or to its Properties of Variation? Variability selection raises an im portan t qu estion abou t adaptive evolu tion : Do adaptation s always reflect a particu lar en viron m en tal state th at dom in ated or arose du rin g an organ ism ’s h istory? Or can th ey reflect th e dyn am ic properties of en viron m en ts su ch as th e ran ge an d tem po of en viron m en tal variation m an ifested over tim e an d space? In variability selection , it is th ese dyn am ic properties th at create th e selective effect. An y en viron m en t a popu lation in h abits m ay in flu en ce th e su rvival an d reprodu ctive su ccess of its m em bers. Wh en prim atologists or an th ropologists do field work, th ey m ay try to evalu ate th e relative com petitive or reprodu ctive su ccess of in dividu als in 94 Evolutionary Anthropology relation to th eir cu rren t en viron m en t. Variability selection raises th e possibility, h owever, th at certain featu res of organ ism s reflect th e past disparity of h abitats rath er th an th e relatively static lan dscape seen over th e sh ort ru n . If evolu tion does in deed in tegrate th e lon g-term effects of h abitat variability, th e n atu re of certain adaptation s m ay be difficu lt to discern in som e livin g organ ism s, if on ly becau se th e observation period is exceedin gly brief com pared to th e tim e fram e of large-scale en viron m en tal rem odelin g. It h as been argu ed 8 th at en viron m en ta l in con sisten cy is r ep etitive a n d th erefore, in a sen se, is a con sisten t elem en t in th e adaptive process of som e organ ism s. E ven by th is argu m en t, adaptive design is n ot en viron m en t-specific bu t is m olded by th e degree an d pace of alteration over tim e. Dram atic flu ctu ation , in an y case, was n ot con sisten t or regu lar du rin g th e Pleistocen e, bu t episodic over m an y gen eration s. Th u s, in evolu tion , tim e m atters: As th e dyn am ic properties of en vir on m en ts u n fold , th e a d a p tive properties of an organ ism m ay be sh aped to m eet episodes of n ovel con dition s experien ced by its lin eage. The Leash of the Past Th e variability-selection h ypoth esis con tradicts certain views th at loom large in th e field of evolu tion ary psych ology. E volu tion ary psych ologists m ain tain th at th e h u m an m in d con sists of m a n y cogn itive sp ecia liza tion s, in clu d in g em otion a l a d a p ta tion s, wh ich r ep r esen t a d eta iled record of th e specific an d recu rren t fea tu r es of Pleistocen e en vir on m en ts.86,87 Th ese specialization s, like all adaptation s, are problem -solvin g devices, an d are said to reflect th e regu larities, u n iform problem s, an d statistical probabilities of h u m an an cestral settin gs. Th ese con sisten t properties of an cestral en viron m en ts are com m on ly kn own as th e en viron m en t of evolu tion ary adaptedn ess.8,10,11 In th e ligh t of ideas I h ave presen ted h ere, th e en viron m en t of evolu tion ary adaptedn ess is a paten tly n arrow con ceptu alization of th e h om in id past. Th e idea was con ceived with ou t an y basis in actu al eviden ce of past en viron m en ts. Som e portraits of th e en viron m en t of evolu tion ary adaptedn ess are m odern h u m an foragin g societies writ in to th e Pleistocen e.88 Th e en vi- ARTI CLES ron m en t of evolu tion ary adaptedn ess an d th e rh etoric of evolu tion ary psych ology, m oreover, affirm th e belief th at com plex adaptation s m u st arise from selective con sisten cy an d u n iform con dition s. E volu tion ary psych ologists com m on ly r efer to ‘‘th e Pleistocen e’’ in th is m on olith ic sen se of h ow an cestral en viron m en ts were su pposedly stru ctu red. As a resu lt, th e h u m an m in d is portr a yed a s a n a m a lga m of sp ecia lpu rpose devices th at reflect a rath er n arrowly con stru ed en velope of adaptive con dition s. Th ere are two m ain problem s with th is view. First, as I h ave sketch ed h ere, th e en velope of con dition s faced by h u m an an cestors was im m en se, with a tren d toward wider extrem es over tim e. H om in id evolu tion was eviden tly in flu en ced by a process of adaptation to th e dyn am ic If evolution does indeed integrate the long-term effects of habitat variability, the nature of certain adaptations may be difficult to discern in some living organisms, if only because the observation period is exceedingly brief . . . an d in con sisten t properties of th ese diverse settin gs. Secon d, evolu tion ary psych ologists seldom distin gu ish th e u n iqu e fu n ction s of h u m an cogn ition from older on es.11 Man y h u m an fu n ction s su ch as fin din g food, m ates, an d sh elter, do in deed reflect u n iversal ch allen ges of su rvival an d reprodu ctive su ccess th at cu t across all en viron m en ts. In an y evolu tion ary an alysis, h owever, it is essen tial to distin gu ish am on g fu n ction s sh ared with m an y an im als, th ose sh ared with oth er h om in ids, an d th ose u n iqu e to Hom o sapien s. Wh ile th e h u m an m in d operates with m an y an cien t m en tal devices, su ch as keen abilities to solve problem s fram ed in a social con text,89 its u n iqu en ess seem s to lie in its effectiven ess in absorbin g n ew, com plex con textu al data an d an alyzin g con dition ality. With in th e perspective of variability selection , it is odd to th in k th at h u m an m en tal life is driven by task-specific devices design ed solely to an alyze an d solve recu rren t Pleistocen e problem s. Wh ile I agree with th e idea of specific circu itry design ed to h an dle adaptive problem s, variability selection wou ld h a ve fa vor ed wid esp r ea d p la sticity with in th e circu itry an d stron g sen sitivity to en viron m en tal in pu t. Th e cogn itive m ech an ism s u n iqu e to h u m an s eviden tly em erged in a com plex series of h igh ly diverse selection regim es. Th ese m ech an ism s th u s in clu de sen sitive collectors an d processors of en viron m en tal data; an alytical devices th at alter beh avior in th e ligh t of com plex con textu al in form ation ; m ediators of n ovel respon se to adaptive problem s; an d calcu lators of tem poral con tin gen cy. Th ese m ech a n ism s p er m it an alysis of factors n ot im m ediately presen t or visible, an d th erefore requ ir e com p lex in ter n a l r ep r esen ta tion , in clu din g lan gu age. Th ese m ech an ism s do n ot com prise som e ‘‘sin gle gen eral-pu rpose learn in g device,’’as den ou n ced by th e evolu tion ary psych ologists. Bu t con siderin g th e n ovelty an d in tricate processin g th ey en able, th ese cogn itive m ech an ism s represen t m ore th an th e h u m an m in d’s tigh t leash to th e m yth ical en viron m en t of evolu tion a r y a d a p ted n ess. Th at is, th ey are m ore th an th e su m of in n ate, special-pu rpose, problem -solvin g devices th at are con stru ed to teth er h u m an cogn ition to th e regu larities of Pleistocen e en viron m en ts.8,86,90 Finding the Boundary Conditions of Hominid Evolution In key tim e span s, h om in ids evolved n ew m ean s of versatility th at en abled th em to exten d th e bou n dary con dition s of th eir evolu tion . Bou n dary con dition s are th e su m of ecological factors th at in flu en ced th e presen ce of h om in ids in a given tim e an d place. For early toolm akers, th ese factors m ay h ave in clu ded th e presen ce of appropriate local rocks for flakin g. For certain species of early au stralopith ecin es, trees with su itable foods or clim bin g su bstrates m ay h ave been part of th eir bou n dary con dition s. Migration s of erectu s-type h om in ids ou t of Africa or H. sapien s to Au stralia an d Evolutionary Anthropology 95 ARTI CLES th e Am ericas reflected an exten sion of bou n dary con dition s. Th at is, adaptive m ech an ism s in som e h om in id popu lation s h ad becom e decou pled from specific local con dition s an d h abitats. Th is decou plin g is an essen tial effect of variability selection . Accordin g to th e variability-selection h ypoth esis, expan sion s of geograph ic ran ge an d en viron m en tal toleran ce were correlated with periods of widen ed disparity in selective con dition s. Testin g th is h ypoth esis will requ ire u s to docu m en t in detail th e ran ge of en viron m en tal variability actu ally en cou n tered by h om in ids an d oth er organ ism s over tim e an d space. As a resu lt, we m ay begin to defin e system atically th e n ovel beh aviors an d ecological factors in volved as h om in ids tracked a favored h abitat or broke particu lar geograph ic barriers. Th u s, th e h ypoth esis sh ou ld h elp to u n cover th e ecological dim en sion s of h om in id evolu tion ary h istory. CONCLUSION H abitat-specific h ypoth eses do n ot adequ ately accou n t for th e m ajor adaptive featu res of h om in id evolu tion . E xtrem e en viron m en tal variability, so am ply docu m en ted in th e geologic record, u ltim ately h ad a m ore pervasive in flu en ce on h om in id evolu tion th an did an y sin gle type of en viron m en t or progressive coolin g or dryin g tren d. Th e va r ia b ility-selection h yp oth esis a ttem pts to ou tlin e a m ech an ism of h ow adaptive evolu tion , in clu din g direction al ch an ge (su ch as en ceph alization ), m ay h ave occu rred in respon se to h abitat variability expressed on a geologic tim e scale. Th e h ypoth esis accou n ts for th e addition , over tim e, of locom otor, cogn itive, an d social m ech an ism s th at m agn ified th e adaptive versatility of certain popu lation s. Th is process u ltim ately affected species’ tu rn over an d th e larger cou rse of h om in id evolu tion . In th e can on of n atu ral-selection th eory, adaptive com plexity evolves as certain organ ism s con sisten tly ou tcom pete oth er organ ism s du e to n ovel alleles or gen e com bin ation s. Th is process, direction al selection , is said to work algorith m ically,3 so th at in crem en ta l ch a n ge a r ises a s a con sequ en ce of con son an t fitn ess resu lts from on e gen eration to an oth er. If th e idea of variability selection is correct, h owever, n ew gen etic possibilities m ay also catch on becau se th ey yield lon gterm win n ers wh ere selective con dition s are in con sisten t over m an y gen eration s. H om in id evolu tion h as produ ced m orph ological an d geograph ic specialists com m itted to certain dietary an d clim atic con dition s. It h as also yielded species th at excel in beh avioral n ovelty, diversity, an d h igh ly soph isticated u se of en viron m en tal data. With its m etabolically expen sive brain an d soph isticated social n etworks, Hom o sapien s is an extrem e exam ple of th e latter type of species. Its evolu tion an d sin gu lar persisten ce are, accordin g to th e variability-selection h ypoth esis, in tim ately lin ked to th e en viron m en tal Habitat-specific hypotheses do not adequately account for the major adaptive features of hominid evolution. in stability th at h as ch aracterized th e h om in id era. ACKNOWLEDGMENTS For th eir com m en ts, su ggestion s, an d argu m en ts, I th an k L.C. Aiello, P. An d r ews, A.K. Beh r en sm eyer, A.S. Br ooks, J.B. Cla r k, J. Da m u th , J. Fleagle, C. Garces, J. Kress, L. Marin o, B. Rich m on d, F. Spoor, D. Strait, E . Vrba, B. Wood, fou r an on ym ou s reviewers, an d oth er colleagu es an d stu den ts from sem in ars at Un iversity College London, the University of Minnesota, U.C. San Diego, and the Sm ithsonian. My gratitude also to Jennifer Clark for assistance with the m anuscript. This is a publication of the Sm ithsonian’s Hum an Origins Program . REFERENCES 1 Dobzh an sky T (1970) Gen etics of the Evolu tion ary Process. New York: Colu m bia Un iversity Press. 2 Mayr E (1970) Popu lation s, S pecies, an d Evolu tion . Cam bridge: H arvard Un iversity Press. 3 Den n ett DC (1995) Darw in ’s Dan gerou s Idea. New York: Sim on & Sch u ster. 4 Lewon tin R (1970) Th e u n its of selection . An n Rev E col Syst 1:1–14. 5 Arn old A, Fristru p K (1982) Th e th eory of evolu tion by n atu ral selection : A h ierarch ical expan sion . Paleobiology 8:113–129. 6 E ldredge N (1989) Macroevolu tion ary Dyn am ics, p 34. New York: McGraw-H ill. 7 Wolpoff MH (1980) Paleoan thropology, pp 21– 22. New York: Kn opf. 8 Tooby J, Cosm ides L (1990) Th e past explain s th e presen t. E th ol Sociobiol 11:375–424. 9 E n dler J (1986) N atu ral S election in the Wild. Mon ographs in Popu lation Biology 21, p 230. Prin ceton : Prin ceton Un iversity Press. 10 Sym on s D (1979) The Evolu tion of Hu m an S exu ality. New York: Oxford Un iversity Press. 11 Foley R (1996) Th e adaptive legacy of h u m an evolu tion : A search for th e en viron m en t of evolu tion ary adaptedn ess. E vol An th ropol 4:194–203. 12 Lewon tin R (1978) Adaptation . Sci Am 239: 156–169. 13 Gou ld SJ, Lewon tin RC (1978) Th e span drels of San Marco an d th e Pan glossian paradigm : A critiqu e of th e adaptation ist program m e. Proc R Soc Lon don 205:581–598. 14 Barth olom ew GA Jr, Birdsell JB (1953) E cology an d th e protoh om in ids. Am An th ropol 55: 481–498. 15 Dart R (1953) Th e predatory tran sition from ape to m an . In t An th ropol Lin gu istic Rev 1:201– 218. 16 Robin son JT (1961) Au stralopith ecin es an d th e origin of m an . An n Rept Sm ith son ian In stit. pp 479–500. 17 Wash bu rn SL (1960) Tools an d h u m an evolu tion . Sci Am 203:63–75. 18 Klein RG (1989) The Hu m an Career. Ch icago: Un iversity of Ch icago Press. 19 Vrba E S (1988) Late Pliocen e clim atic even ts an d h om in id evolu tion . In Grin e F (ed), Evolu tion ary History of the ‘‘R obu st’’ Au stralopithecin es, pp 405–426. New York: Aldin e de Gru yter. 20 Pren tice ML, Den ton GH (1988) Th e deep-sea oxygen isotope record, th e global ice sh eet system an d h om in id evolu tion . In Grin e F (ed), Evolu tion ary History of the ‘‘R obu st’’ Au stralopithecin es, pp 383–403. New York: Aldin e de Gru yter. 21 deMen ocal PB (1995) Plio-Pleistocen e African clim ate. Scien ce 270:53–59. 22 Vrba E S, Den ton GH , Partridge TC, Bu rkle LH (eds) (1995) Paleoclim ate an d Evolu tion w ith Em phasis on Hu m an Origin s. New H aven : Yale Un iversity Press. 23 Cerlin g TE , H arris JM, MacFadden BJ, Leakey MG, Qu ade J, E isen m an n V, E h lerin ger JR (1997) Global vegetation ch an ge th rou gh th e Miocen e/ Pliocen e bou n dary. Natu re 389:153–158. 24 Vrba E S, Den ton GH , Pren tice ML (1989) Clim atic in flu en ces on early h om in id beh avior. Ossa 14:127–156. 25 Stan ley S (1992) An ecological th eory for th e origin of Hom o. Paleobiology 18:237–257. 26 Cerlin g TE (1992) Developm en t of grasslan ds an d savan n as in E ast Africa du rin g th e Neogen e. Palaeogeogr Palaeclim atol Palaeoecol 97:241– 247. 27 Cerlin g TE , Qu ade J, Am brose SH , Sikes NE (1991) Miocen e fossil soils, grasses an d carbon isotopes from Fort Tern an (Ken ya): Grasslan d or woodlan d? J H u m E vol 21:295–306. 28 Kin gston JD, Marin o BD, H ill A (1994) Isotopic eviden ce for Neogen e h om in id paleoen viron m en ts in th e Ken ya Rift Valley. Scien ce 264:955– 959. 29 Sikes N (1994) E arly h om in id h abitat preferen ces in E ast Africa: Paleosol carbon isotopic eviden ce. J H u m E vol 27:25–45. 30 Wh ite TD, Su wa G, Asfaw B (1994) Au stralopithecu s ram idu s, a n ew species of early h om in id from Aram is, E th iopia. Natu re 371:306–312. 31 Wh ite TD, Su wa G, Asfaw B (1995) Au stralopithecu s ram idu s, a n ew species of early h om in id from Aram is, E th iopia. Corrigen du m . Natu re 375:88. 96 Evolutionary Anthropology 32 WoldeGabriel G, Wh ite TD, Su wa G, Ren n e P, de H ein zelin J, H art WK, H eiken G (1994) E cologica l tem p or a l p la cem en t of ea r ly Pliocen e h om in ids from Aram is, E th iopia. Natu re 371:330– 333. 33 Clarke RJ, Tobias PV (1995) Sterkfon tein Mem ber 2 foot bon es of th e oldest Sou th African h om in id. Scien ce 269:521–524. 34 Bu rger L (person al com m u n ication ). 35 Rayn er RJ, Moon BP, Masters JC (1993) Th e Makapan sgat au stralopith ecin e en viron m en t. J H u m E vol 24:219–231. 36 Boesch -Ach erm an n H , Boesch C (1994) H om in ization in th e rain forest: Th e ch im pan zee’s piece of th e pu zzle. E vol An th ropol 3:9–16. 37 Blu m en sch in e RJ (1986) Early Hom in id S caven gin g Opportu n ities: Im plication s of Carcass Availability in the S eren geti an d N goron goro Ecosystem s. Oxford: British Arch aeological Reports In tern ation al Series 283. 38 Cavallo JA, Blu m en sch in e RJ (1989) Treestored leopard kills: E xpan din g th e h om in id scaven gin g n ich e. J H u m E vol 18:393–399. 39 Potts R (1996) Hu m an ity’s Descen t: The Con sequ en ces of E cological In stability. New Yor k: William Morrow. 40 Potts R (1996) E volu tion an d clim ate variability. Scien ce 273:922–923. 41 deMen ocal PB, Bloem en dal J (1995) PlioPleistocen e clim atic variability in su btropical Africa an d th e paleoen viron m en t of h om in id evolu tion . In Vrba E S, Den ton GH , Partridge TC, Bu rckle, LH (eds), Paleoclim ate an d Evolu tion w ith Em phasis on Hu m an Origin s, pp 262–288. New H aven : Yale Un iversity Press. 42 Ku kla G (1989) Lon g con tin en tal records of clim ate—an in trodu ction . Palaeogeogr Palaeoclim atol Palaeoecol 72:1–9. 43 Traverse A (1982) Respon se of world vegetation to Neogen e tecton ic an d clim atic even ts. Alch erin ga 6:197–209. 44 Gu iot J, Pon s A, de Beau lieu J-L, Reille M (1989) A 140,000-yr con tin en tal clim ate recon stru ction from two E u ropean pollen records. Natu re 338:309–313. 45 Ku kla G, Zh ish en g An (1989) Loess stratigraph y in cen tral Ch in a. Palaeogeogr Palaeoclim atol Palaeoecol 72:203–225. 46 Roberts N (1984) Pleistocen e en viron m en ts in tim e an d space. In Foley R (ed), Hom in id Evolu tion an d Com m u n ity Ecology, pp 25–53. Lon don : Academ ic Press. 47 Rossign ol-Strick M (1983) African m on soon s, an d im m ediate clim ate respon se to orbital in solation . Natu re 304:46–49. 48 Taylor KC, Lam orey GW, Doyle GA, Alley RB, Grootes PM, Mayewski PA, Wh ite JWC, Barlow LK (1993) Th e ‘‘flickerin g switch ’’ of late Pleistocen e clim ate ch an ge. Natu re 361:432–436. 49 MacDon ald GM, E dwards TWD, Moser KA, Pien itz R, Sm ol JP (1993) Rapid respon se of treelin e vegetation an d lakes to past clim ate warm in g. Natu re 361:243–246. 50 Dorale JA, Gon zález LA, Reagan MK, Pickett DA, Mu r r ell MT, Ba ker RG (1992) A h igh resolu tion record of H olocen e clim ate ch an ge in speleoth em calcite from Cold Water Cave, n orth east Iowa. Scien ce 258:1626–1630. 51 Su tcliffe AJ (1985) On the Tracks of Ice Age Mam m als. Cam bridge: H arvard Un iversity Press. 52 Clem en s SC, Mu rray DW, Prell WL (1996) Non station ary ph ase of th e Plio-Pleistocen e Asian Mon soon . Scien ce 274:943–948. 53 Paytan A, Kastn er M, Ch avez FP (1996) Glacial to in terglacial flu ctu ation in produ ctivity in th e equ atorial Pacific as in dicated by m arin e barite. Scien ce 274:1355–1357. ARTI CLES 54 deMen ocal PB, Ru ddim an WF, Pokras E M (1993) In flu en ces of h igh - an d low-latitu de processes on African terrestrial clim ate. Palaeocean ograph y 8:209–242. 55 Feibel CS (1997) Debatin g th e en viron m en tal factors in h om in id evolu tion . GSA Today 7:1–7. 56 Feibel CS, H arris JM, Brown FH (1991) Neogen e paleoen viron m en ts of th e Tu rkan a basin . In H arris JM (ed), Koobi Fora R esearch Project, vol. 3, pp 321–370. Oxford: Claren don Press. 57 Potts R (1994) Variables versu s m odels of early Pleistocen e h om in id lan d u se. J H u m E vol 27:7–24. 58 Raym o ME , Ru ddim an WF (1992) Tecton ic forcin g of late Cen ozoic clim ate. Natu re 359:117– 122. 59 West-E berh ard MJ (1989) Ph en otypic plasticity an d th e origin s of diversity. An n Rev E col Syst 20:249–278. 60 Travis J (1994) E valu atin g th e adaptive role of m orph ological plasticity. In Wain wrigh t PC, Reilly SM (eds), Ecological Morphology, pp 99–122. Ch icago: Un iversity of Ch icago Press. 61 Dein o A, Potts R (1990) Sin gle crystal 40 Ar/39 Ar datin g of th e Olorgesailie Form ation , sou th ern Ken ya rift. J Geoph ys Res 95:8453–8470. 6 2 Potts R, Dein o A (1995) Mid -Pleistocen e ch an ge in large m am m al fau n as of th e sou th ern Ken ya rift. Qu atern ary Res 43:106–113. 63 Gin sberg J (1989) In Jewell PA, Maloiy GMO (eds), The Biology of Large African Mam m als in Their Environm ent. London: Clarendon. 64 Owen RB, Ren au t RW (1981) Palaeoen viron m en ts an d sedim en tology of th e Middle Pleistocen e Olorgesailie Form ation , Sou th ern Ken ya Rift Valley. In Coetzee J, van Zin deren Bakker (eds), Palaeoecology of Africa, pp 147–174. Rotterdam : Balkem a. 65 Potts R (n.d.) Environm ental hypotheses of hom inin evolution. Yearbook Phys Anthropol, in press. 66 Kaplan RH , Cooper WS (1984) Th e evolu tion of developm en tal plasticity in reprodu ctive ch aracteristics: An application of th e ‘‘adaptive coin flippin g’’ prin ciple. Am Nat 123:393–410. 67 West-E berh ard MJ (1986) Altern ative adaptation s, speciation , an d ph ylogen y (a review). Proc Natl Acad Sci USA 83:1388–1392. 68 Slatkin M, Lan de R (1976) Nich e width in a flu ctu atin g en viron m en t—den sity in depen den t m odel. Am Nat 110:31–55. 69 Bu ll JJ (1987) E volu tion of ph en otypic varian ce. E volu tion 41:303–315. 70 Real LA (1980) Fitn ess, u n certain ty, an d th e role of diversification in evolu tion an d beh avior. Am Nat 115:623–638. 71 Clark AB (1991) In dividu al variation in respon siven ess to en viron m en tal ch an ge. In Box H O (ed), Prim ate R espon ses to En viron m en tal Chan ge, pp 91–110. Lon don : Ch apm an an d H all. 72 Kidwell SM, Beh ren sm eyer AK (1993) Taph on om ic approach es to tim e resolu tion in fossil assem blages. S hort Cou rses in Paleon tology, N o. 6. Tu lsa: Th e Paleon tology Society. 73 Beh ren sm eyer AK, H ook RW (1992) Paleoen viron m en tal con texts an d taph on om ic m odes. In Beh ren sm eyer AK, Dam u th JD, DiMich ele WA, Potts R, Su es H -D, Win g SL (eds), Terrestrial Ecosystem s Throu gh Tim e, pp 15–136. Ch icago: Un iversity of Ch icago. 74 Levin s R (1968) Evolu tion in Chan gin g En viron m en ts. Prin ceton : Prin ceton Un iversity Press. 75 Raff RA (1996) The S hape of Life: Gen es, Developm en t, an d the Evolu tion of An im al Form . Ch icago: Un iversity of Ch icago Press. 76 H ollin g CS (1973) Resilien ce an d stability of ecological system s. An n Rev E col Syst 4:1–23. 77 Wilson E O, Bossert WH (1971) A Prim er of Popu lation Biology. Su n derlan d, MA: Sin au er Associates. 7 8 Potts R (1991) Wh y th e Old owa n ? PlioPleistocen e toolm akin g an d th e tran sport of resou rces. J An th ropol Res 47:153–176. 79 Potts R (n .d.) Beh avioral respon se to variable Pleistocen e lan dscapes. In Nowell A (ed), The Archaeology of In telligen ce. New York: Cam bridge Un iversity Press (in press). 80 H ay RL (1976) Geology of the Oldu vai Gorge. Berkeley: Un iversity of Californ ia Press. 81 Rogers MJ, Feibel CS, H arris JWK (1994) Ch an gin g pattern s of lan d u se by Plio-Pleistocen e h om in ids in th e Lake Tu rkan a Basin . J H u m E vol 27:139–158. 82 Mellars P (1995) The N ean derthal Legacy. Prin ceton : Prin ceton Un iversity Press. 83 Ru ff CB, Trin kau s E , H olliday TW (1997) Body m ass an d en ceph alization in Pleistocen e Hom o. Natu re 387:173–176. 84 Aiello LC, Wh eeler P (1995) Th e expen sivetissu e h ypoth esis: Th e brain an d th e digestive system in h u m an an d prim ate evolu tion . Cu rr An th ropol 36:199–221. 85 Ru ff CB, Walker A (1993) Body size an d body sh ape. In Walker A, Leakey R (eds), The N ariokotom e Hom o erectu s S keleton , pp 234–265. Cam bridge: H arvard Un iversity Press. 86 Tooby J, Cosm ides L (1989) E volu tion ary psych ology an d th e gen eration of cu ltu re, part I. E th ol Sociobiol 10:29–49. 87 Tooby J, Cosm ides L (1993) Th e psych ological fou n dation s of cu ltu re. In Barkow JH , Cosm ides L, Tooby J (eds), The Adapted Min d: Evolu tion ary Psychology an d the Gen eration of Cu ltu re, pp 19–136. New York: Oxford Un iversity Press. 88 Ridley M (1993) The R ed Qu een . New York: Macm illan . 89 Cosm ides L (1989) Th e logic of social exch an ge: H as n atu ral selection sh aped h ow h u m an s reason ? Stu dies with th e Wason selection test. Cogn ition 31:187–276. 90 Pin ker S (1994) The Lan gu age In stin ct. New York: William Morrow. 91 Wrigh t S (1932) Th e roles of m u tation , in breedin g, crossbreedin g an d selection in evolu tion . Proc Sixth In t Con gress Gen et 1:356–366. 92 Wrigh t S (1968–78) Evolu tion an d the Gen etics of Popu lation s, vols. 1–4. Ch icago: Un iversity of Ch icago Press. 93 Fish er RA (1958) The Gen etical Theory of N atu ral S election , 2n d ed. New York: Dover. 94 Provin e WB (1986) S ew ell Wright an d Evolu tionary Biology. Chicago: University of Chicago Press. 95 Sh ackleton NJ (1995) New data on th e evolu tion of Pliocen e clim atic variability. In Vrba E S, Den ton GH , Partridge TC, Bu rckle LH (eds), Paleoclim ate an d Evolu tion w ith Em phasis on Hu m an Origin s, pp 242–248. New H aven : Yale Un iversity Press. 96 Pren tice ML, Den ton GH (1988) Th e deep-sea oxygen isotope record, th e global ice sh eet system an d h om in id evolu tion . In Grin e FE (ed), Evolu tion ary History of the ‘‘R obu st’’ Au stralopithecin es, pp 383–403. New York: Aldin e de Gru yter. 97 Woodru ff F, Savin SM, Dou glas RG (1981) Miocen e stable isotope record: A detailed deep Pacific Ocean stu dy an d its paleoclim atic im plication s. Scien ce 212:665–668. 98 Wrigh t JD, Miller KG (1992) Miocen e stable isotope stratigraph y, Site 747, Kergu elen Plateau . Proc Ocean Drillin g Program , Sci Resu lts 120: 855–866. 99 Miller KG, Mou n tain GS, et al. (1996) Drillin g an d datin g New Jersey Oligocen e-Miocen e sequ en ces: Ice volu m e, global sea level, an d E xxon records. Scien ce 271:1092–1095. r 1998 Wiley-Liss, Inc.