Mary C. StinerDepartment of Anthropology,University of Arizona,Tucson, AZ 85721-0030,U.S.A. E-mail:mstiner@u.arizona.edu

Received 7 October 1996Revision received 25 March1997 and accepted12 October 1997

Keywords:mortality analysis,Ursus, Spelearctos, cavebears, brown bears, blackbears, Pleistocene cavefaunas, Paleolithic,hibernation, hunting,scavenging.

Mortality analysis of Pleistocene bears andits paleoanthropological relevance

Bear bones and Paleolithic stone artefacts often co-occur inPleistocene cave deposits of Eurasia, raising the question of how theseassociations come about and the need for effective methods withwhich to obtain a clear answer. Building upon knowledge of modernbears, I present a method for testing two competing hypotheses aboutthe causes of bear mortality in hibernation contexts. The firsthypothesis proposes that age-dependent deaths resulted from non-violent causes (principally starvation), implying that bears’ presencein a cave was not linked in time to human activities there. The secondhypothesis proposes that random bear deaths in caves resulted fromhunting by humans or other large predators, implying a temporal linkbetween them; the expectation of a nonselective age pattern in thiscircumstance arises from the fact that the individual characters ofhibernating bears are hidden from predators. Three elements of themethod and its development are presented: (1) a brief review of thebiological bases of hibernation-related mortality in modern Ursus, itspaleontological consequences, and test expectations drawn there-from; (2) a detailed, illustrated technique for age-scoring isolatedbear cheek teeth based on tooth eruption-wear sequences, developedprimarily for cave and brown bears; and, (3) a simple, accurate way toevaluate real cases in terms of contrasting mortality models. The finalstep is demonstrated by application to a Middle Pleistocene cave bearassemblage (Ursus deningeri) from Yarimburgaz Cave in Turkey,a large collection found in general stratigraphic association withPaleolithic artefacts. The advantages of the method include its abilityto (a) handle small samples, (b) use isolated tooth specimens, and (c)evaluate cases simultaneously in terms of idealized age structuremodels and the variation that normally is associated with each undernatural conditions. While the more obvious benefit of bear mortalityanalysis may be to research on ancient bear demography, the prin-ciples and procedures offered here are equally pertinent to archaeo-logical studies of carnivore-mediated formation processes in cavesites. As is generally true in taphonomic research, however, bearmortality patterns are most effective when used in combination withindependent lines of evidence to address questions about assemblageformation.

? 1998 Academic Press Limited

Journal of Human Evolution (1998) 34, 303–326

Introduction

Bear skeletal remains are encountered withgreat regularity in Pleistocene cave faunasof Eurasia, including many Paleolithicarchaeological sites (e.g., Baryshnikov,1997; Chauvet et al., 1996; Gargett, 1996;Jenkinson, 1984; Kurtén, 1976; Pratt, 1988;Stiner, 1994; Stiner et al., 1996). The pres-

0047–2484/98/030303+24 $25.00/0/hu970198

ence of bear skulls and postcrania in sheltersites—sometimes in association with stonetools, other times not—has been taken bysome as evidence of archaic human hunting,magic, or ritual behavior, interpretationsinspired in part by the prominence of bearsin the iconographies of some Holocenehuman cultures. Scenarios of Neandertal‘‘cave bear cults’’, and forced eviction by

? 1998 Academic Press Limited

304 . .

spearpoint of giant bears from cave men’shomes, continue to hold tremendouspopular appeal. Yet claims for evidence ofarchaic human rituals, in which bear skulls(Bächler, 1940, 1957 cited in Kurtén 1976),deer antlers (Blanc, 1958, 1961; Piperno,1976–77), or certain other non-human skel-etal materials are cited as central props orsymbols, are highly problematic (Binford,1981; Chase & Dibble, 1987; Gamble,1986; Kurtén, 1958, 1976; Spahni, 1954;Stiner, 1991, 1994; White & Toth, 1991).Cases dating to the dawn of the humanspecies, well before any unequivocal form ofart is preserved in archaeological records,are notoriously difficult to verify. There issurprisingly little evidence of direct inter-action between early humans and bearsoverall.Archaeologists must admit, however, that

their abilities to falsify fanciful scenariosabout archaic human behavior remainunderdeveloped. Bear-artefact associationsin Pleistocene caves are points of specialcontention. On the one side, doubters notethat the bones of cave and brown bears inPaleolithic shelter sites seldom are marredby stone tools. Indeed, many professionalsare now convinced that bear remains inPaleolithic cave sites most likely representhibernation-related deaths that occurredindependently of human uses of the samelocalities. Kurtén’s (e.g., 1958, 1976) inves-tigations have been particularly influential inthis regard, due to his effective use of refer-ence data on modern ursids and innovativeanalyses of paleontological samples. Tapho-nomic work on other Pleistocene cases (e.g.,Brain, 1981; Jenkinson, 1984; Stiner, 1994;Stiner et al., 1996) also demonstrates thatthe tempos at which geological and biologi-cal materials accumulate in caves can bequite different. Where cave sediments accu-mulate slowly, diverse biological and non-biological agencies may contribute materialsto a single geological stratum. This is why somany cave faunas represent palimpsests or

‘‘overlays’’ of multiple bone accumulationevents, such as in Yarimburgaz Cave innorthwest Turkey (Stiner et al., 1996),selected data on which are presented below.On the other hand, a number of observa-

tions can be marshaled in favor of directinteractions between Pleistocene humansand bears, or at least isolate weaknesses inthe opposing view. It is generally known,after all, that humans can in some circum-stances process game without leaving tracesof tool marks on prey skeletons (e.g.,Lyman, 1994), and occasional predation onhibernating bears by humans is widely docu-mented among modern traditional cultureswhere these species’ geographic ranges over-lap (e.g., Rogers, 1981:69; Ross et al.,1988). Tool marks occasionally have beenfound on bear bones in a few Paleolithiccases dating to the Middle Paleolithicand onward (e.g., Bárta, 1989; Stiner1994:109ff). Hence, it continues to be diffi-cult to distinguish hominids’ foraging refusein cave sites from non-anthropogenic faunalcomponents, even if we are confident abouthow such processes may operate in prin-ciple. Obviously, some clear-cut negationstrategy is required on a case by case basis.There are numerous benefits to learning

why bears are common in Paleolithic cavesites, even if the conclusion proves to be thatbears avoided humans as best they couldwhen selecting winter beds. Knowledgeabout the occupation intervals of bears andhumans in caves, for example, helpsdescribe the ecological conditions underwhich human visits took place. Bear mor-tality patterns represent one independentand potentially pivotal line of evidence ofbear denning (hibernation) behavior(Andrews & Turner, 1992; Baryshnikov,1997; Gargett, 1996; Kurtén, 1958, 1976;Stiner, 1994; Stiner et al., 1996), a potentialexplanation of why Pleistocene bear bonesoccur in caves, sometimes in great numbers.Left open, however, is the question of howbears die in Paleolithic caves. Hinging on

305

this question are interpretations of directphysical contact versus indirect ecologicalinteractions between humans and bears,and, in some cases, whether unusualmanipulation of skeletal materials by archaichumans took place.Reported here is a method for analyzing

the age structures of bear death assem-blages, designed to address the question ofwhether Pleistocene bear remains found incaves represent hibernation-related deathsarising from non-violent (predator-independent) or violent (predator-dependent) causes. The first of these alter-natives argues for temporal independenceof bear and human components in caves.Because it represents the most likely non-anthropological explanation in these cir-cumstances, it must be excluded beforeanthropological explanations may be for-warded. Of course, bear mortality data aremost valuable to research on site formationhistories if used in conjunction with inde-pendent lines of taphonomic evidence suchas vertebrate species profiles, bear body partrepresentation, and bone damage patterns(e.g., Gargett, 1996; Stiner et al., 1996;Wolverton, 1996): a preponderance ofbears suggests a den accumulation, forexample; whole skeletons of bears also arguefor deaths in situ; and bone damage mayreveal who, if anyone, gained from thosedeaths.Several studies have documented how

approximate age at death in bears may beinferred from the stages of dental develop-ment or occlusal wear (Dittrich, 1960;Kurtén, 1958, 1976; Marks & Erickson,1966; Torres, 1988), as well as how the agestructure of a bear death assemblage maybe constructed from these observations(Andrews & Turner, 1992; Gargett, 1996;Kurtén, 1958). With the exception ofKurtén’s (1958) life table analysis, whichfocuses on age-specific survivorship inextinct cave bears, none of these studies hastaken what I consider to be the final steps

toward interpretation of bear mortality pat-terns. Clear, divisible test implications areessential for comparing subject cases againstalternative age structure models, especiallythose representing common mortality fac-tors in nature (sensu Caughley, 1966, 1977;Lyman, 1987; Stiner, 1990). Because somestructural variation is typical of every model(Caughley, 1977), the possibility that a caseresembles alternative models needs to befalsified.Age structure models are narrow, static

representations of what is actually a dynamicprocess. Conceiving of the models as ‘‘pat-tern families’’, based on empirically definedranges of variation, is a simple way of copingwith this shortcoming (Stiner, 1990, 1991,1994). Both considerations—idealizedstructural models and the variation sur-rounding each—are needed to distinguishbetween conflicting interpretations of bearremains in cave sites, such as attritionalmortality in hibernation dens arising fromstarvation and disease as opposed to theconsequences of in situ predation (surpriseattacks in dens) by Pleistocene humans,hyenas, wolves, and/or other bears.Herein lie the main contributions of this

study. The test expectations are of tapho-nomic and behavioral interest because theyare explicit and founded on independentlyformulated demographic principles com-mon to mammals. Additional benefits of theapproach include its fully illustrated agingscheme, which coordinates the eruption andwear schedules for six different cheek teeth,and its ability to handle small samples effec-tively. The age classification techniqueis equally suited to samples consistingof isolated or articulated teeth. Thispresentation is slanted to archaeologicalresearch on site formation processes, but theapproach is also of value to paleoecologicalstudies of Pleistocene brown and cavebears.In order to appreciate the clarity of

the test expectations for bear mortality in

306 . .

cave sites, it is necessary to begin withsome demographic and reproductivecharacteristics that modern bears hold incommon, but which may not be obviousto anthropologists. Development of themethod therefore involves three steps thatare pertinent to hypothesis formation, datacollection, and interpretation of results.First to be explored are the causal relationsamong bear hibernation behavior, life his-tory characteristics, and mortality in andaround den localities. A common theme isevident in the biology of modern membersof the genus Ursus (black bear, U. ameri-canus; brown bear, U. arctos; polar bear,U. maritimus), suggesting that many of theirreproductive and metabolic characteristicsalso apply to extinct Middle and LatePleistocene cave bears [the chronospeciesUrsus (Spelearctos) deningeri and U. (S.)spelaeus, as well as U. (S.) rossicus]. Theseuniversal characteristics therefore serve asthe basis for a series of test expectations.Second, an age-scoring technique is pre-sented, based on eruption and occlusal wearsequences for the six major cheek teeth ofbears over the maximum potential lifespan.Finally, a means for evaluating observedmortality patterns in relation to classicdemographic models and their associatedranges of variation is presented. This thirdstep is illustrated by application to a realbear death assemblage found in geologicalassociation with Paleolithic stone tools inYarimburgaz Cave. Results obtained for sixtypes of cheek teeth are checked for levels ofagreement in this context.

Bear mortality in hibernation contexts:test implications

Knowledge of hibernation behavior and itscontingencies in modern bears permitsdevelopment of expectations for age struc-ture patterns arising in frequently used densfrom attritional mortality and from in situ

predation. At issue are two questions: Whatfraction of individuals in a living populationnormally hibernate?—this serving as themodel for random predation by den raiders.And which age groups are most commonlyaffected by hibernation-related mortalityarising primarily from non-violent causesincluding starvation, disease, and sen-escence? It should be noted at the outsetthat modern bears are known to use naturalshelters—including rock crevices, over-hangs, and true caves—as hibernation densif they are available. This behavior is es-pecially common where bear habitats areassociated with limestone karst systems(Clevenger, 1991; Judd et al., 1986;Reynolds et al., 1976; Rogers, 1981); well-drained localities hidden by tree cover andwith moderately steep taluses are preferred.Den sites may be reused by bears, andpermanent shelters are likely to attractdifferent individuals over time.

Life history characteristics and hibernationHibernation by bears is a means for enduringthe scarcity of plants, invertebrates, andsmall vertebrates in winter (e.g., Clevenger,et al., 1992; Ewer, 1973; Garshelis & Pelton,1980; Rogers, 1981, 1987; Tassi, 1983).Rather than depending on large game inseasons when other foods are no longeravailable, bears possess the metabolic optionto drastically reduce their need for energy(Folk et al., 1976; Hellgren et al., 1990;Johnson & Pelton, 1980; Rogers, 1981;Watts & Jonkel, 1988; Watts et al., 1987).The success of the hibernation strategyhinges, however, upon food availability dur-ing the previous warm season, especiallyautumn. Mortality in modern bears there-fore tends to be cyclical, and mortality isespecially high toward the end of the hiber-nation period in stressful years (Garshelis &Pelton, 1980; Kurtén, 1976; Rogers, 1981,1987). Starvation is the most common causeof hibernation-related deaths in most studyareas, especially for cubs and subadults

(pfims1yaby(1bma(cbdccmreaarpfda1dmeronih

gwPtT1

307

(Craighead et al., 1976; Rogers 1987).Starving adults generally do not die in sleep:they awaken early if their energy stores areover-depleted and make short forays insearch of easy food, using the den as a forag-ing hub (Rogers, 1987:24–25, 70). Bearsmay hang about the den for up to a monthafter first emergence, and unfortunate onescollapse and die in the vicinity.Hibernation lairs are also birthing dens for

pregnant females. Delayed implantation offertilized zygotes in adult females permitsspontaneous adjustments to food supply andsynchronization of births to the hibernationschedule (Bunnell & Tait, 1981; Bronson,1989; Ewer, 1973:298–300). Zygotes can beaborted at little biological cost if the femaleenters hibernation in poor condition.Delayed implantation also maximizes thedevelopment time available to healthy butnaturally altricial young before abandoningthe den in spring.Reproducing females tend to den longer

than other adults in a given population(Rogers, 1981). Cubs of both sexes willhibernate with their mothers each winteruntil she rejects their company, normallybetween the second and fourth year inbrown bears. The duration of cub depen-dency also varies within and between popu-lations as a function of foraging conditions.Cubs that are abandoned early may formsibling duos, inseparable until they reach 2·5to 3 years of age (Glenn et al., 1976),indirectly testifying to the hazards of earlyindependence. The importance and dur-ation of hibernation for adult males andbarren females may be the same or less thanthat of reproducing females, depending onthe severity of seasonal lows in plant andsmall animal productivity and whethernon-reproducing bears can obtain a regularsupply of meat from large game in winter(Johnson & Pelton, 1980; Rogers, 1987).Key aspects of bear life history and winter

torpor are remarkably similar amongmodern species of Ursus. Bunnell & Tait’s

1981) comparisons of black, brown, andolar bears reveal nearly analogous ages atrst parturition (between 4–8 years for all),aximum potential lifespans (25–30 years,ee also Craighead et al., 1976; Glenn et al.,976), average birth intervals (2 to >3ears), and litter sizes (1–2 in most studyreas). The cubs of brown and polar bearsecome self-sufficient between 1·5 and 4·5ears of age, black bears somewhat earlier0·5 to 2·5 years) (Bunnell & Tait,981:82). Bears therefore are characterizedy long lifespans, also relatively late sexualaturation, protracted reproductive cycles,nd high survivorship rates in adulthoodCraighead et al., 1976). Bears evidentlyan afford a low reproductive capacityecause of their large body size, effectiveefenses against predators, high maternalompetence, and occasionally cooperativeub care (cub switching) among neighboringothers. The average reproductive rateeported for grizzly bears in the Yellowstonecosystem by Craighead et al. (1976; seelso Ewer, 1973), for example, is 0·658 pernnum, meaning that even minor changes ineproductive rate can affect the potential foropulation growth or recovery; a 25 year oldemale may achieve roughly six full repro-uctive cycles over her reproductive careernd, in that time, expect to produce about3 cubs. Bears have among the lowest repro-uctive rates known among terrestrial mam-als (Bunnell & Tait, 1981:77; Craigheadt al., 1976; Glenn et al., 1976) and, for thiseason, modern bears are very sensitive tover-exploitation by hunters. It therefore isot surprising that bear meat forms only anrregular source of food in any modernuman economy.The cave bear’s apparently strong phylo-enetic affinities to modern bears, alongith their paleontological abundance inleistocene cave sediments, indicate thathey too were hibernators (Andrews &urner, 1992; Baryshnikov, 1997; Kurtén,958, 1976). Similarities in the hibernation

308 . .

process and demography of modern blackbears (U. americanus, Hellgren et al., 1990;Garshelis & Pelton, 1980; Johnson & Pelton,1980; Rogers, 1981, 1987) and brown bears(U. arctos, for southern European browns,see Clevenger, 1990, 1991; Clevenger et al.,1987, 1988; Clevenger & Purroy, 1991;Tassi, 1983) therefore permit certain gener-alizations about all bears (Watts & Jonkel,1988; Watts et al., 1987; see also Hellgrenet al., 1990; Johnson & Pelton, 1980),including Holocene and Pleistocene brownand cave bears.Food supply is arguably the most import-

ant limiter of bear population sizes inmodern environments because it directlyimpacts birth rates. Maternal nutritionlimits a females’ reproductive rate bothprior to zygote implantation and duringinfant development before and followingbirth (Bunnell & Tait, 1981:83–84;Craighead et al., 1976:354; see also Picton& Knight, 1986). Mortality factors play asecondary role in bear population dynamics,largely because the death rate in adults isquite low prior to senescence. Mortalityrates are substantially higher for cubsunder 1·5 years of age, however: up to 38%of the litter may be lost during this time,but losses ranging from 0 and 15% aremore typical (Glenn et al., 1976:389).Many of these deaths occur in the vicinity ofnatal dens. Kurtén (1958:25–27) obtainedsimilar estimates of post-partum juvenilemortality rates for Pleistocene cave bearsfrom Odessa, along with an apparentlylow probability of death during the primeadult years. Because cave bears appear tohave been highly dependent on seasonallyavailable plant mast, based on studies ofdental morphology and stable isotope analy-ses of tooth enamel (e.g., Kurtén, 1976;Baryshnikov, 1997; Stiner et al., 1998),and omnivorous diets extend hibernationtimes for non-reproducing adults, one canexpect relatively long hibernation periodsfor adults of both sexes. The net result

is a nearly equal probability for adults ofboth sexes perishing in dens (Stiner et al.,1998).In addition to non-violent causes of death,

denning bears may be attacked by predatorssuch as wolves, dogs, humans, or otherbears. A bear becomes nearly helpless as itdrops into a reduced metabolic state. It is forthis reason that bears make considerableeffort to keep their hibernation lairs secret(for black bears, see Garshelis & Pelton,1980; Johnson & Pelton, 1980; Rogers,1981, 1987; for brown bears see Clevenger,1991; Miller, 1985; Peterson et al., 1984:38;Rogers, 1987; Ross et al., 1988). Any sign ofdisturbance at a potential winter lair site isenough to turn a bear away. Cannibalismwithin and between bear species at dens andelsewhere can be commonplace during leanyears, or in regions where bear populationdensities are especially high (Mattson et al.,1992; Ross et al., 1988; Tietje et al., 1986).From a taphonomic viewpoint it should benoted that both scavenging and predation ofhibernating bears can result in gnawingdamage to bones. Frail bears may triggerattacks by other predators, but the carcassesof already dead bears would likely bescavenged.Hibernation dens clearly are magnets

for annual mortality in bears. The circum-stances of den use greatly enhance the prob-ability that bear bones will become part ofthe paleontological records of natural shel-ters, especially if the shelter is frequentedacross generations. However, bears can diefrom more than one set of causes in hiber-nation contexts—essentially violent andnonviolent—even if nonviolent causes pre-dominate in modern situations. The uniquebiology of bears brings about distinctmortality patterns in dens in predominantlyviolent and predominantly nonviolent situ-ations. These distinct age structures havedifferent implications for bears’ connectionto prehistoric human activities at thesame sites.

309

Expectations for cumulative deaths inhibernation densIn light of the information presented above,four facts about bear life history and forag-ing habits appear to be especially importantfor modeling expectations of cumulativemortality in hibernation contexts:(1) Most bear deaths in hibernation dens

are caused by starvation, disease, and oldage.(2) Bears produce very small litters per

reproductive event.(3) Cub mortality rates are moderate to

high, whereas those for adults are quite lowprior to senescence.(4) Nearly the entire bear population

(adults and juveniles) can be found in densfor substantial portions of each year if thebears are highly omnivorous, a qualificationwhich almost certainly includes cave bears,most brown bears, and black bears. Polarbears are totally carnivorous, and they arethe exception that proves the rule linkinghibernation times, diet, and sex ratios indens.These observations lead to the prediction

that young and very old adult bears shouldbe preferentially represented in hibernation-related death assemblages where starvation,disease, and senescence are the primarymortality factors. Prime adult bearsshould be conspicuously under-representedbecause their survivorship is high prior tosenescence. In histogram format, such amortality pattern would display a bimodal,or U-shaped (concave), age distribution.When compared to the age structure ofa living bear population (Figure 1), themortality profile resulting from normal non-violent attrition (the NNVA model) is dis-tinguished foremost by the near absence ofprime-aged adults. In habitats where foodsupplies allow males to hibernate for lesstime, females may predominate among theadults that spend long periods in dens. Theproportion of juveniles to adults also may beamplified slightly as a result, but not to a

great degree since bears produce cubs inlow numbers over long intervals. For thesereasons bear mortality patterns in densshould not usually result in a juvenile biasstronger than that predicted by the NNVAmodel, because the small litter sizes of bearsdamp this tendency relative to most otherden-dependent Carnivora (cf. strong juven-ile biases for wolves and hyenas, Stiner,1994:316–331; see also Kurtén’s generaldiscussion, 1958:4–5).The mortality pattern expected to result

from non-violent deaths in dens (NNVA)contrasts with that from hunting of hiber-nating bears by predators. Surprise attackson sleeping bears by modern wolves andhumans are well known, though relativelyuncommon (see references above);Pleistocene hyenas occasionally may havedone the same when their distributions over-lapped with those of bears. If the practicewas more common in the past, the mortalitypattern generated by repeated hunting ofhibernating bears should be indifferent tothe ages of the individuals occupying dens,simply because there is little opportunity tospook or survey individual vulnerabilitywhile hunting encrypted prey. Repeatedpredation on hibernating bears in a cavetherefore should affect prime adults, oldadults, infants, and adolescents randomly,emulating their natural proportions in theliving population sequestered in denseach year. Here, then, is the alternativeexpectation, namely the living structure(LS) model.Although cut marks on Pleistocene bear

bones are relatively rare in archaeologicalsites overall, gnawing damage by largecarnivores frequently is noted (Andrews &Turner, 1992; Baryshnikov, 1997; Gargett,1996; Kurtén, 1976; Stiner et al., 1996;see also Wolverton, 1996 for a Holoceneexample). These observations raise thequestion of how the bears died but do notnecessarily answer it. Gnawing damage onbear remains—or cut marks in the case

310 . .

of human exploitation—is not proof ofhunting, only consumption or relatedeconomic uses. It is the mortality patternthat potentially allows one to distinguishthe circumstances of bone damage,specifically between hunting and scavengingcontexts.

The age-scoring technique

This technique for constructing bear mor-tality patterns employs teeth exclusively, andit is suitable for age-scoring isolated speci-mens. Ideally, juveniles and adults also can

be distinguished on the basis of bone fusion(e.g., Marks & Erickson, 1966), but thetaphonomic contexts of cave faunas, specifi-cally the prevalence of gnawing damage byvarious large carnivores, often eclipses thispossibility. The situation is different fordental remains, including deciduousteeth, providing that the enamel caps arenearly or fully formed at the time of death(following Wainwright et al., 1976:223–224). Teeth present an additional advantagefor age structure analyses in that theyundergo continuous, highly diagnosticalterations over the complete lifetime, first

ixiAge cohorts

n in

divi

dual

s

viii iii iv v vii viii

Figure 1. An idealized attritional (U-shaped or bimodal) mortality profile, indicated by vertical bars,superimposed upon the idealized model (grid) of the living population age structure. Under-representation of individuals in the prime adult age cohorts is the most important difference between thetwo idealized models. While a U-shaped mortality pattern can arise from non-violent causes or from openland predation by cursorial predators (see Stiner, 1990, 1994), only normal non-violent attrition (NNVA)applies in the case of hibernation-related deaths of bears. In situ predation on hibernating bears shouldinstead affect individuals randomly with respect to age. In situ predation therefore should result in a deathpattern that resembles the living age structure (LS) of the hibernating population, which in most bearsclosely resembles the structure of the entire living population. This kind of graphic representation doesnot, however, reflect the structural variation typically associated with either idealized model.

311

via development and later through occlusalwear of the crown.1

1Ways of age-scoring teeth of modern and wellpreserved fossil populations include eruption-wear andcementum annulus counting techniques (e.g., Marks& Erickson, 1966; Stoneberg & Jonkel, 1966). Whilebeneficial to wildlife research, the applicability ofcementum annulus counting is limited for paleontologi-cal populations both by the poor preservation potentialof these fragile connective tissues and the technique’sdestructive nature. By contrast, evidence of tooth erup-tion and wear is preserved as long as the tooth elementitself remains recognizable, and therefore is theapproach to age-scoring favored here. The nondestruc-tive aspect of eruption-wear approaches also allowsmore than one investigator the opportunity tostudy various dental aspects of a population, as well asto evaluate prior findings on mortality patterns inparticular.

In developing an eruption and wearscheme for bear teeth, it is relevant to noteursids’ common adaptation to an omnivo-rous diet and the tendency for the occludingsurfaces to be large in relation to crownheight. The only exception to these general-izations is the recent dental specialization ofthe polar bear (U. maritimus). Pleistocenecave bears evolved toward another extreme,with massive, broad, multi-cusped den-titions for processing plant foods with toughstems and husks (Baryshnikov, 1997;Kurtén, 1958, 1976; Stiner et al., 1998).Despite some dental variation among livingand extinct bear species, the cheek teeth ofall bears can be classified into consecutiveage categories (stages) on the basis of crownand root development or, if mature, wear ofoccluding surfaces. Here I emphasize thestatus of permanent tooth crowns.Figures 2 and 3 describe the age-scoring

technique as a set of nine consecutivecohorts for upper and lower cheek teeth ofbears. The scheme is developed foremost forcave bears but can be applied with minimaladjustments to other bear species, especiallybrown bears. Unlike any other age-scoringdiagram published to date, the synchronizederuption and wear stages illustrated here arekeyed to the quantitatively based age struc-ture models and their respective test impli-

cations (see below). In this regard, it isanalytically self-contained. This techniquedoes not attempt, however, to estimate agein real years for any age stage, nor does itassume that the stages represent equivalentspans of time, as neither is really possiblefor prehistoric populations on the basis oferuption and wear observations. Thescheme registers change in wear status on anordinal scale.The age-scoring technique was developed

with the benefit of relatively completemaxillae and mandibles of cave and brownbears from Middle Pleistocene levels ofYarimburgaz Cave in Turkey, and fromUpper Pleistocene levels in Buca della Iena,Grotta del Fossellone, Grotta dei Moscerini,and Grotta di Sant’Agostino in Italy (Stiner,1994). The paleontological observations aresupplemented by dental development dataon modern brown and black bears (Dittrich,1960; Marks & Erickson, 1966; Torres,1988). The age-scoring technique empha-sizes the molars and, to a lesser extent, theupper fourth premolar. Canines are not con-sidered because their patterns of wear anddamage were found to be very irregular inthe Pleistocene reference collections (Stineret al., 1998; for other cases see Koby, 1940,1953), and among some modern brownbears (Clevenger, personal communication,1994). Moreover, permanent caninesemerge late in the development sequence(Marks & Erickson, 1966; Dittrich, 1960),and the deciduous elements that precedethem generally do not register much in theway of measurable wear (but see Andrews &Turner, 1992).An important parameter for the eruption-

wear scheme is the maximum potentiallifespan (MPL). As noted above, the MPLcannot be determined in real years on thebasis of occlusal wear for paleontologicalpopulations. Instead, complete destructionof cheek tooth crowns, especially thosesituated in the area of maximum wear, sig-nals the end of the MPL on grounds that the

312 . .

animal’s food processing apparatus isseverely compromised or no longer func-tional (Stiner, 1990, 1994). The anteriormolars may be best for age-scoring, becausethey emerge fairly early, they experienceheaviest friction overall, and completedestruction of the tooth crown is possiblewith advanced age (Figures 2 and 3).Deciduous teeth are de-emphasized by this

scheme in favor of partly developed perma-nent teeth, because of the nominal structureof the deciduous elements and their shortuse-life in relation to the time at which cubsabandon the winter den to follow theirmothers.Because the bear tooth eruption-wear

scheme is developed from real speci-mens, some variation between individuals,

Figure 2. The tooth eruption wear stages for upper cheek teeth of cave and brown bears (genus Ursus),based on the P4, M1, and M2 (right side). The age stages are divided into nine cohorts, collapsible to three.The sequence begins with development of the deciduous tooth and/or the permanent counterpart (both instage I, see text), followed by successive categories of occlusal wear; the arrow signifies that the permanenttooth crown is still emerging. The sequence ends (stage IX) with partial or complete destruction of toothcrown and exposure of the root pulp cavity, depending on the position of the element in the dental row.The M1 represents the zone of greatest cumulative wear, and those for the P4 and the M2 are adjusted toit. Note that the scheme does not attempt to estimate real age for any cohort, nor does it assume that thecohorts are of equal duration; neither assumption is necessary to the age structure comparisons. Drawingsare of real, representative teeth and, for this reason, the wear patterns observed in other assemblages cannot be expected to display perfect matches; the total area of enamel polish or dentin exposure is moreimportant than the details of enamel loss.

313

populations, and species is to be expected,both in the shapes of dentin patches exposedby occlusal wear and in the absolute corre-spondences between the onset of wear andcompletion of root development in youngindividuals (see also Dittrich, 1960:133–135). Inter-individual variation in the degreeof tooth wear in mammals naturallyincreases with age (e.g., Lowe, 1967; Murie,1951). It also is likely that bear populationsoccupying distinct habitats experiencesomewhat different wear rates as a functionof dietary fiber and grit (cf. Andrews &Turner, 1992; Baryshnikov, 1997; Koby,1940, 1953; Kurtén, 1976:18ff; Stiner et al.,1998). However, these are relatively minor

sources of ambiguity in this empiricallyderived nine-stage scheme and do notappear to be a problem for the simplerthree-age scheme presented below.

Scoring system for nine age cohortsPermanent cheek tooth eruption sequencesin modern Ursus (following Dittrich, 1960;Kurtén, 1958; Marks & Erickson, 1966) arerelatively uniform. Crown developmentbegins with the upper and lower first molars(M1), followed by fourth premolars (P4),second molars (M2), lower third molars(M3), and finally the canines (C) (e.g.,Figure 4). Eruption of lower third molar andcanine crowns occurs during the second

Figure 3. The tooth eruption wear stages for lower cheek teeth of cave and brown bears (genus Ursus),based on the M1, M2, and M3 (left side). The age stages are divided into nine cohorts, collapsible to three.The sequence begins with development of the deciduous tooth and/or the permanent counterpart (both instage I, see text), followed by successive categories of occlusal wear; the arrow signifies that the permanenttooth crown is still emerging. The sequence ends (stage IX) with partial or complete destruction of thetooth crown and exposure of the root pulp cavity, depending on the position of the element in the dentalrow. The M1 and M2 represent the zone of greatest cumulative wear, and wear for the M3 is adjusted tothem. Note that the scheme does not attempt to estimate real age for any cohort, nor does it assume thatthe cohorts are of equal duration; neither assumption is necessary to the age structure comparisons.Drawings are of real, representative teeth and, for this reason, the wear patterns observed in otherassemblages can not be expected to display perfect matches; the total area of enamel polish or dentinexposure is more important than the details of enamel loss. The cusp morphology and wear patterns of theM3 appear to be especially variable among individuals.

314 . .

winter of life in modern bears; all of thepermanent teeth with the exception of thecanines will have pushed through the gumsby end of first year, however. Permanentcrown development normally is completedwithin the second year of life; root develop-ment is completed much later (e.g., Poelker& Hartwell, 1973).Age stage I in Figures 2 and 3 potentially

combines counts for deciduous upper orlower canines with those for emerging butunworn permanent teeth. Brown bear cubsdevelop the full deciduous dentition only bythe third month of life, and the first perma-nent molars (M1) emerge in the fifth month.Infant bears are especially altricial amongcarnivores (Ewer, 1973:296, 332), and theirmilk teeth are unimpressive (Dittrich, 1960;Kurtén, 1958). Because the deciduous den-tition of bears is less elaborate than those ofcanids or hyaenids, some qualifications tothe definition of the juvenile versus adultcohorts in bears are required. Specifically,an analyst’s decision to include deciduousteeth in stage I must take into account the

fact that their presence in a living animaloverlaps with erupting (countable) perma-nent tooth crowns for significant spans oftime. Because deciduous teeth may straddleemerging permanent ones in a young bear,Stage I must include all permanent teethfrom the germ stage to those with rootlengths that are up to 50% complete, butonly those deciduous teeth that display nowear to moderate wear. This procedure mayslightly underestimate the total number ofvery young individuals for an assemblage,but it avoids the more serious problem ofdouble-counting these individuals (Stiner,1994:322–324).Stage II is represented by a fully (or

nearly) emerged permanent tooth crown,whose root development exceeds 50% ofthe normal length at maturity but whoseocclusal surface lacks visible wear. Littlewear normally occurs until after rootformation is completed (95–100%).Thus, occlusal wear stages I–II include allphases of root development for permanentcheek teeth.

Figure 4. Juvenile (male?, specimen 90-P-306) and adult (female, specimen 90-P-288) cave bearmandibles from Yarimburgaz Cave. Note the encrypted status and vertical position of the lower thirdmolar during crown formation. At the same age, eruption of the second molar is nearly completed, but thecanine crowns are only about 10% erupted. The posterior part of the tooth row ‘‘unfolds’’ as the growingjaw lengthens.

315

By stage III, root development is com-pleted and some wear of the occlusal enamelis evident, but little or no dentin is exposed. Abear that dies during stage II or III wouldstill be young, probably hibernating along-side its mother through a second or even athird winter. Stages IV through IX are basedexclusively on the wear status of the occlusalsurface. The reference assemblages used todevelop the scheme reveal that the M1, M2,and M1 can be worn down to the roots bystage IX in exceptionally long-lived individ-uals, often with the pulp cavities exposed.Age-scoring of the posterior molars (M3,M2) therefore requires adjustments relativeto the anterior molars to compensate fortheir ‘‘younger’’ appearance (see Figures 2and 3).

Collapsing nine cohorts to three forcomparisons to models in tripolar formatPhrased initially in terms of nine age stages,the eruption-wear diagrams in Figures 2 and3 include guidelines for collapsing the datainto three age stages—juvenile (I–III), primeadult (IV–VII), and old adult (VIII–IX). Thisis the most efficient way to examine theresults in terms of modeled variation andthe test implications outlined above. Anadditional advantage to this simpler schemeis its tolerance for small samples. As few as12 individuals may be acceptable, althoughlarger samples are preferred as a rule; thisminimum threshold is considerably lowerthan that required for nine-cohort analysis,minimally 30 animals according to Lyman(1987).The three simplified age stages—juvenile,

prime adult, and old adult—effectivelydescribe the essential differences among theage structure models in question. This isbecause the three age groups are dividedaccording to major changes in physiologyand behavior in the life histories of femalemammals (see Stiner, 1990, 1994). Theplacements of the age divisions are import-ant. The juvenile-prime adult juncture

compensates for the relatively precociouspattern of tooth development in carnivoresrelative to most herbivores by relegatingslightly worn permanent teeth to the juvenilecategory. The transition to adulthood corre-sponds to the time at which females are firstable to reproduce, regardless of whetherthey have learned enough to be competentparents. The prime-old adult age boundaryis defined by a significant decline inadult female fertility, normally at about67% of MPL. In modern brown bearsthis begins soon after 15 years of age,although cubs have been successfullywhelped by female bears in the 22nd yearof life (Craighead et al., 1976:346–350).The onset of the reproductive decline inadult females therefore fits well with the67%-of-MPL boundary established pre-viously by Stiner (1990, 1994) for othermammals.Figure 5(a) and (b) show how mortality

data organized into the three age categoriescan be compared to a series of mortalitymodels and their attendant variation (‘‘pat-tern families’’) in a tripolar format. Thegraph consists of three age axes [Figure5(a)], each bisecting the triangle and rang-ing from 0 to 100%. By converting thefrequencies of individuals in each agecategory to percentages of the total, the agestructure based on a particular tooth ele-ment and population can be represented bya single point. Results for different toothelements of the same population, or resultsbased on one type of tooth for several popu-lations, can be compared as multiple pointson a single graph.Figure 5(b) presents the interpretative

standards—or models—for the mortalityanalysis. The areal distributions of fivemortality pattern families are based on thereal consequences of a host of mortalityfactors, including disease, accidents, mal-nutrition, and/or predation (from Stiner,1990, 1994). Most important to thisdiscussion are the two areas in the lower

316 . .

central zone of the graph, labeled normalnon-violent attrition (NNVA) and living-structure (LS) patterns respectively (follow-ing Caughley, 1977 on mammals; for bearsspecifically, see Rogers, 1987; Bunnell &Tait, 1981; Eberhardt et al., 1986;McCullough, 1981). Each corner of thegraph represents a significant bias toward

the designated age group. As explainedearlier, the prediction for bear assemblagesrepresenting hibernation deaths is theNNVA pattern type. The prediction forbear death assemblages representingprimarily random predation in situ (e.g.,by wolves, hyenas, or humans) is the LSpattern type.

Figure 5. Guides for reading (a) the tripolar graphs and (b) the pattern map representing modeledvariation associated with five distinct families of mortality patterns (from Stiner, 1990). Each arearepresents a mortality pattern family, incorporating both the average condition and normal variationsurrounding it. Of greatest relevance to interpreting bear mortality patterns in cave sites are the normalnon-violent attrition (NNVA) pattern family and the living structure (LS) pattern family. The areaidentified as NNVA (*) overlaps with mortality patterns generated in prey caught in open land conditionsby cursorial predators (see Stiner, 1990, 1994), but this is not relevant to interpretation of hibernation-related mortality patterns in bears. In situ predation on den-bound bears should instead affect individualsrandomly with respect to age and therefore result in a death pattern that resembles the living age structure(LS) of the hibernating population.

317

Application to the Middle Pleistocenecave bear assemblage from

Yarimburgaz Cave

The large bear death assemblage fromYarimburgaz Cave in northwestern Turkey(Arsebük et al., 1990, 1991, n.d.; Arsebuk &Özbasaran, 1994; Howell & Arsebük, 1989,1990; Özdogan & Koyunlu, 1986; Stineret al., 1996) was used in part to develop thiseruption-wear approach to mortality analy-sis, along with several smaller assemblagesfrom Italian cave sites. Central to under-standing the formation history of the MiddlePleistocene faunas in Yarimburgaz Cave isthe question of why bears and Paleolithicartefacts occur in the same geological levels.Application to the Yarimburgaz sampleillustrates the potential of the method fortaphonomic research on other cases inwhich a temporal link between human andbear occupations is possible.Yarimburgaz Cave consists of two inter-

nally connected chambers, the lower ofwhich extends roughly 600 m into the lime-stone bedrock. Middle Pleistocene sedi-ments (Blackwell et al., 1990; Farrand,1992) excavated within 70 m of the lowerentrance are rich in both cave bear remains(Stiner et al., 1996) and Paleolithic artefacts(Kuhn et al., 1996). Bear bones constitute93% of all macromammal specimens recov-ered from the Middle Pleistocene deposits;ungulate and non-ursid carnivore bonesmake up the remaining 4% and 3% respect-ively. Taphonomic investigations by Stiner,Arsebük & Howell (1996) indicate thatthe bone and artefact accumulations repre-sent a palimpsest of multiple, discreteoccupations of the cave by bears andhominids. Bear mortality data are amongthe important kinds of evidence pointing tothis conclusion.As many as 45 individual bears are repre-

sented in the bone assemblage, most ofthem a large variant of Ursus [Spelearctos]deningeri, accompanied by rare examples of a

small-bodied brown bear (U. arctos) (Stineret al., 1998). The skeletal remains are in arelatively advanced state of fossilization. Themacroscopic features of the bones and teethare very well preserved, as are traces ofgnawing by various carnivores and, in otherinstances, small rodents (see Stiner et al.,1996). Skeletal representation for the bearsverges on anatomical completeness, alsoconsistent with the hypothesis that the bearremains accumulated in the cave as theresult of many hibernation-related deathevents.The complete taphonomic evaluation of

the Yarimburgaz case is in fact based oncross-referenced information on bone modi-fication, species and body part represen-tation, and mortality analysis, syntheticresults on which are reported elsewhere(Stiner et al., 1996). One arm of that inves-tigation, focusing on bear mortality patterns,shows that the bear remains are primarilythe result of non-violent mortality in hiber-nation dens (the NNVA model), notpredator-caused mortality (the LS model).A result indicating predominantly non-violent mortality negates the possibility oftemporal connections between humans’ andbears’ presence in the cave.Table 1 presents age frequency data in

nine-cohort format for the cave bears ofYarimburgaz, based on each of six differentcheek tooth elements (P4, M1, M2, M1, M2,and M3). All excavation units are combinedin this comparison because the question isabout the net consequences of multiple bearoccupation and death events. A U-shaped,or bimodal, contour emerges in all of thehistograms shown in Figure 6, with thepossible exception of that for the P4. Twodistinct peaks occur in cohorts III and VIII,whereas prime-aged adult bears are under-represented. To this point, the resultsobtained for the Yarimburgaz cave bears arebroadly consistent with those obtained byKurtén (1958) for cave bears from Odessa,Gargett (1996) for Pod Hradem in the

318 . .

Table1

Nine-cohortmortalitydataforYarimburgaz

cave

bearsbased

onsixdifferentcheektoothelem

ents

Tooth

element

Num

berof

age-scored

teeth

MNI

Nineagecohorts:

III

III

IVV

VI

VII

VIII

IX

P4

3719

77

95

53

01

0M

143

258

611

52

01

82

M2

3620

79

82

20

16

1M

165

355

620

137

51

44

M2

7944

912

2713

53

27

1M

362

397

1315

74

43

72

Note:Thenumberofage-scoredteethincludesrightandleftelements.Theminimum

numberofindividualanimals(M

NI)isbasedon

themostcommon

side

(rightorleft);itisnotused

foranyofthecalculationstofollowbutinsteadisprovided

asbackground

information.

319

Czech Republic, Wiszniowska (1982) forBacho Kiro in Bulgaria, and Andrews &Turner (1992) for cave bears fromWestburyin Great Britain. Possible overlap of theobserved patterns with other mortalitymodels (especially LS) has not been refuted,however (see below).A nine-cohort approach also reveals some

interesting details about juvenile mortality inthe cave bears. Stage III juveniles are espe-cially prevalent. Two factors may contributeto this result. First, cubs younger than threemonths may not possess erupted teeth, andthe partly formed deciduous tooth crownsdeveloping below the gumline are not forti-fied by dentin and therefore may be moreprone to destruction. A few very young cubsin Yarimburgaz Cave may have been com-pletely ingested by predators, evidenced by acanid scat sample containing the bones of ahare and an infant bear (Stiner et al., 1996).The low proportion of individual bears incohort I could therefore be due partly toa low probability of preservation (see alsoAndrews & Turner, 1992:145–147). How-ever, this explanation can not apply tocohort II, which also contains fewer individ-uals than cohort III. The cheek teeth ofcohort II individuals would have been well-formed at the time of death and therefore asresistant to decomposition factors as theteeth representing cohort III.The second factor contributing to the

high proportion of individuals in cohort IIImay be heightened mortality for this latejuvenile age stage. Indeed, the pattern forthe Yarimburgaz cave bears is consistentwith the situation among modern brownbears of North America (see, for example,Bunnell & Tait, 1981; Glenn et al., 1976):the peak in cohort III likely represents thepivotal winter in which some juveniles foundthemselves alone for the first time since birthand many were unable to survive this criticaltransition period. The fact that adult femalebears can spontaneously shed fertilized,unimplanted zygotes in years when the

foraging conditions prior to hibernation arepoor may serve as a ‘‘bottleneck’’ beyondwhich only those cubs whose survivalchances are very good make it to parturition.Modern bears are attentive mothers, foster-ing relatively high survivorship in cubs whileunder their care. Life may be far moredifficult, by contrast, during the year inwhich young bears must begin to fend forthemselves. Those abandoned early are atgreatest risk.Although the mortality histograms for the

six tooth elements in Figure 6 could be saidto differ from one another, a Kolmogorov–Smirnov test (Table 2) comparing thedistributions shows that they yield fairlyconsistent results. Only three of the 15 pair-wise age profile comparisons indicate differ-ences at the 0·05 level of probability. Thedifferences always involve upper-lower ele-ment pairings [Table 2(b)]; least agreementis found between the profiles for the M2 andM1. The K-S test suggests—but does notprove—that all or most of the histogramsare assignable to a single mortality patternfamily.The results presented thus far say nothing

about how the findings relate to alternativemortality models—a question of what theobserved patterns are not. In recognition ofthe variation that inevitably surrounds eachidealized age structure model, the final stepevaluates the extent to which a real case fitsone model as opposed to all others. Thisprocedure is very awkward in a two-dimensional format and often proves to beguesswork. It is relatively easy to obtainrigorous results by using the tripolar formatoutlined in Figure 5.When the Yarimburgaz cave bear data are

collapsed into the three age cohorts (Figure7 and Table 3), it becomes clear that all ofthe age structures indicated by the six cheekteeth fulfill the NNVA prediction. The nine-cohort histograms vary in their contours(Figure 6), but not enough to warrant theirclassification as substantially different types

320 . .

IX

10

0I

Age cohorts

N

VI

5

III IV V VII VIIIII

M2

10

0

N 5

P4

10

0

N 5

M1

10

0

Age cohorts

N 5

M3

10

0

N 5

M1

10

0

N 5

M2

IXI VIIII IV V VII VIIIII

321

of mortality pattern (Figure 7). The upperand lower molars of the bears form separatesubgroups on the tripolar graph, but all yieldessentially the same answer.The NNVA mortality pattern of the

Yarimburgaz cave bears is consistent withhibernation-related deaths arising primarilyfrom starvation and disease. This mortalitypattern, which affects mostly young andold individuals, differs significantly fromrandomly-selected age groups normallyfound in hibernation dens. The NNVAmor-tality pattern of the Yarimburgaz cave bearstherefore can not be explained as resultingprimarily from surprise attacks by hominids,wolves or hyenas. What damage occurs on

the bear remains is attributable to non-ursidlarge carnivores (wolves and/or hyenas) andoccasional cannibalism by other bears(Stiner et al., 1996; see also Gargett, 1996on Pod Hradem). Hunted bears cannot bedistinguished from scavenged ones on thebasis of bone damage, but the mortalityevidence strongly argues for deaths primarilyfrom starvation and other non-violent attri-tional factors, and that the remains of thebears were subsequently scavenged by pred-ators. While some bear bones exhibit tracesof gnawing, no cut marks were found. Itseems that humans were not among thescavengers that utilized bear carcasses inYarimburgaz Cave.

Figure 6. Age structure histograms for the Yarimburgaz bear death population based on six different cheektooth elements and nine age stages. All of the histograms display U-shaped, bimodal contours in thistwo-dimensional format due to a relative paucity of prime-aged adults, with the possible exception of thatfor the P4. The profiles exhibit subtle variation among tooth elements. Juveniles in cohort III are especiallyprevalent in most profiles, possibly representing deaths following/during the first winter of independence.N represents the average number of individual bears, calculated by summing right and left sides of a giventooth element and dividing the sum by two.

Kolmogorov–Smirnov two sample test results of a comparison of mortalityprofiles for the Yarimburgaz cave bears, based on six different cheek toothelements(a) K-S values:

M1 M2 M3 M1 M2 P4

M1 0·0M2 0·131 0·0M3 0·163 0·107 0·0M1 0·238 0·170 0·112 0·0M2 0·366 0·290 0·208 0·151 0·0P4 0·306 0·233 0·175 0·225 0·194 0·0

(b) two-sided probability values:

M1 M2 M3 M1 M2 P4

M1 1·000M2 0·581 1·000M3 0·380 0·835 1·000M1 0·106 0·393 0·877 1·000M2 0·0031 0·028* 0·261 0·722 1·000P4 0·0211 0·112 0·435 0·228 0·452 1·000

*Differences that are significant at the 0.05 level of probability.

Table 2

322 . .

DiscussionAlthough the more obvious benefit of mor-tality data on Pleistocene bears may be inthe area of bear paleoecology, the mortalitypatterns of bears in caves are also importantto paleoanthropological research. A review

of how modern bears live renders bear–artefact associations in Pleistocene cavesremarkable in relation to modern human–bear interactions. The fact that any cavecould have been frequented repeatedlyby bears, Paleolithic hominids or other

Figure 7. Three-age results for six different cheek teeth of Yarimburgaz bears graphed in tripolar format.The distribution shows that the mortality data for the Yarimburgaz cave bears fall within the NNVApattern family; each of the six cheek tooth elements indicates a classically attritional age structure for thedeath population, consistent with the expectation of non-violent mortality in hibernation dens. Thepossibility that the mortality pattern of the Yarimburgaz cave bears resembles the living structure (LSpattern family) of the hibernating population is falsified. (This figure is reprinted with permissionfrom ‘‘Cave bears and Paleolithic artefacts in Yarimburgaz Cave, Turkey: dissecting a palimpsest’’, byM. C. Stiner, G. Arsebük, and F. C. Howell, Geoarchaeology copyrighted 1996, Academic Press.)

Mortality data for Yarimburgaz bears, using three age cohorts

Toothelement

Number ofage-scoredteeth

Percentages of total:

JuvenilePrimeadult

Oldadult

P4 37 62% 35% 3%M1 43 58 19 23M2 36 67 14 19M1 65 48 40 12M2 79 61 29 10M3 62 56 29 14

Table 3

323

carnivores suggests that human use of theseplaces was relatively ephemeral and/or thatoccupations were separated by long inter-vals. Hibernation deaths in bears areseasonal phenomena and might occur whenhuman occupation of the same place wasleast likely. However, any sign of recentdisruption at a shelter can be enough todissuade a den-seeking bear—their greatvulnerability during dormancy requires thatthe hibernation lair be kept secret. Thecommon presence of bear skeletal remainsand stone artefacts in Paleolithic caves ofEurasia therefore suggests extended win-dows of safe use for bears. It is likely underthese circumstances that the occupations byhumans and bears were widely scattered intime, on the order of decades or longer, andthat claims of the shelter by any prospectiveoccupant hinged primarily upon a low risk ofinterference or disturbance.Mortality patterns add a unique and valu-

able dimension to research on the circum-stances of bear–artefact associations inPaleolithic shelter deposits. The age struc-ture of Middle Pleistocene cave bears fromYarimburgaz Cave falls squarely into theNNVA pattern family and is exactly whatone would expect to result from primarilynon-violent attrition in a hibernation dencontext. Mortality data therefore areamong the kinds of evidence that negate thepossibility of any direct causal (temporal)link between bear and hominid occupationsof Yarimburgaz Cave, arguing insteadthat the bears used the cave privately as ahibernation lair over many generations.The boundaries dividing the three agestages provide a relatively coarse descriptionof age structure, yet they prove effectivefor taphonomic investigations of carnivorepresence in Paleolithic caves (see alsoStiner, 1994).It is, however, the combination of mul-

tiple independent sources of informationthat constitutes a compelling argument forhibernation deaths as the main or sole cause

of the bear bone accumulations in the cave.The mortality data are simply one highlydesirable line of evidence. Of course thevalue of mortality analysis is contingent onhow it is applied. The method presentedhere fosters replicability of results and theirinterpretation across studies by pushinganalysis beyond simply choosing an expla-nation to testing it. The method has thepower to isolate contrasting outcomes,because the age divisions used are closelyallied to major transitions in the life course,and the reference models are developedfrom independent empirical and theoreticalsources. Needless disagreements amongresearchers have occurred in the recentzooarchaeological literature simply forthe lack of independent standards forconstructing, classifying, and interpretingage structure profiles.

Acknowledgements

I thank F. Clark Howell and Güven Arsebükfor inviting me to participate in theYarimburgaz project. I am also very gratefulto Mihriban Özbasaran, Mehmet Özdogan,Sevil Gülçur, Gönül Egeli, Oguz Tanindi,Nur Balkan-Atli, Korkut Atli, and others fortheir logistical help and good company dur-ing my work at Istanbul University. Alsoimportant to this research were the accom-modations provided by the AmericanResearch Institute in Turkey (ARIT,Istanbul) and practical advice from AnthonyGreenwood (ARIT Director) and LeylâNisli (Secretary). The data collection phaseof this research, which focused primarilyon the anthropological implications of theYarimburgaz Cave faunas, was supportedby a grant from the Wenner-Gren Foun-dation. Comments by Steve Kuhn, R. LeeLyman, Leslie Aiello, Peter Andrews, BlaireVan Valkenburgh, Scott Wing, and anony-mous reviewers were invaluable for thedevelopment of this manuscript.

324 . .

References

Andrews, P. & Turner, A. (1992). Life and death of theWestbury bears. Ann. Zool. Fennici 28, 139–149.

Arsebük, G., Howell, F. C. & Özbasaran, M. (1990).Yarimburgaz 1988. XI. Kazi Sonuclari Toplantisi, I.Ankara Ûniversitesi Basimevi, pp. 9–38.

Arsebük, G., Howell, F. C. & Özbasaran, M. (1991).Yarimburgaz 1989. XII. Kazi Sonuclari Toplantisi, I.Ankara Ûniversitesi Basimevi, pp. 17–41.

Arsebük, G., Farrand, W., Howell, F. C., Kuhn, S.,Özbasaran, M., Schwarcz, H. P., Stiner, M. C.,Tsoukala, E. & Von Koenigswald, W. (n.d.). Yarim-burgaz Cave, Final Report. Unpublished documentsubmitted to the Wenner-Gren Foundation.

Arsebük, G. & Özbasaran, M. (1994). Yarimburgazmagaralari: Pleistosen’den bir kesit. XI Türk TarihKongresi. Türk Tarih Kurumu Basimevi, Ankara,1994, pp. 17–27.

Bächler, E. (1940). Das Alpine Paläolithikum derSchweiz. Basel.

Bächler, E. (1957). Die Altersfliederung der Höhlen-bärenreste im Wildkirchli, Wildenmannisloch undDrachenloch. Quartär 9, 131–146.

Bárta, J. (1989). Hunting of brown bears in the Meso-lithic: evidence from the Medvedia Cave near Ruzínin Slovakia. In (C. Bonsall, Ed.) The Mesolithic inEurope, pp. 456–460. Edinburgh: John Donald.

Baryshnikov, G. (1997). Cave bears from the Paleo-lithic of the Greater Caucasus. In (J. J. Saunders,B. W. Styles & G. Baryshnikov, Eds) QuaternaryPaleozoology in the Northern Hemisphere. ScientificPapers of the Illinois State Museum, Springfield(in press).

Binford, L. R. (1981). Bones: Ancient Men and ModernMyths. New York: Academic Press.

Blackwell, B., Schwarcz, H. P., Porat, N., Howell,F. C. & Arsebük, F. (1990). Electron spin resonance(ESR) dating of Ursus teeth from Yarimburgaz Cave,Turkey. Abstracts of the Geological Society of America1990 22(3201), A120–121.

Blanc, A. C. (1958). Torre in Pietra, Saccopastore,Monte Circeo, on the position of the Mousterianin the Pleistocene sequence of the Rome area. In(G. H. R. Von Koenigswald, Ed.) Hundert JahreNeanderthaler (Neanderthal Centenary), pp. 167–174.Utrecht: Wenner-Gren Foundation.

Blanc, A. C. (1961). Some evidence for the ideologiesof early man. In (S. L. Washburn, Ed.) The SocialLife of Early Man, pp. 119–136. Chicago: Aldine.

Brain, C. K. (1981). The Hunters or the Hunted?Chicago: University of Chicago Press.

Bronson, F. H. (1989). Mammalian ReproductiveBiology. Chicago: University of Chicago Press.

Bunnell, F. L. & Tait, D. E. N. (1981). Populationdynamics of bears—implications. In (C. W. Fowler &T. D. Smith, Eds) Dynamics of Large Mammal Popu-lations, pp. 75–98. New York: John Wiley & Sons.

Caughley, G. (1966). Mortality patterns in mammals.Ecology 47, 906–917.

Caughley, G. (1977). Analysis of Vertebrate Populations.London: John Wiley & Sons.

Chase, P. G. & Dibble, H. (1987). Middle Paleolithicsymbolism: a review of current evidence and inter-pretations. Journal of Anthropolological Archaeology 6,263–293.

Chauvet, J.-M., Deschamps, É. B. & Hillaire, C.(1996). Dawn of Art: the Chauvet Cave. New York:Harry N. Abrams.

Clevenger, A. (1990). Biologia del Oso Pardo (Ursusarctos) en la Cordillera Cantabrica. Memoria de TesisDoctoral, Universidad de Leon.

Clevenger, A. (1991). The phantom bear of theSpanish Sierras. Wildlife Conservation 94, 34–45.

Clevenger, A. P. & Purroy, F. J. (1991). Ecologia del OsoPardo en Espana. Monographias del Museo Nacionalde Ciencias Naturales, Madrid.

Clevenger, A., Purroy, F. J. & Saenz de Buruaga, M.(1987). Status of the brown bear in the CantabrianMountains, Spain. International Conference on BearResearch and Management 7, 1–8.

Clevenger, A., Purroy, F. J. & Pelton, M. R. (1992).Food habits of brown bears (Ursus arctos) in theCantabrian Moutains, Spain. J. Mammal. 73, 415–421.

Craighead, J. J., Craighead, F. C. & Sumner, J. (1976).Reproductive cycles and rates in the grizzly bear,Ursus arctos horribilis, of the Yellowstone ecosystem.In (M. R. Pelton, J. W. Lentfer & G. Edgar Folk,Eds) Bears—Their Biology and Management, Third(1974) International Conference on Bear Research andManagement, pp. 337–356. Morges: InternationalUnion for Conservation of Nature, ICUN NewSeries, no. 40.

Dittrich, L. (1960). Milchgebißentwicklung und Zahn-wechsel beim Braunbären (Ursus arctos L.) undanderen Ursiden (Dissertation). Morph. Jahrb.101(1), 1–141.

Eberhardt, L. L., Knight, R. R. & Blanchard, B. M.(1986). Monitoring grizzly bear population trends.Journal of Wildlife Management 50, 613–618.

Ewer, R. F. (1973). The Carnivores. London:Weidenfeld & Nicholson.

Farrand, W. R. (1992). Geoarchaeology of Yarimbur-gaz Cave, Turkey. Evolución Humana en Europa y losYacimientos de la Sierra de Atapuerca, Actas, Vol. 1.Junta de Castilla y León, pp. 19–36.

Folk, G. E. Jr, Larson, A. & Folk, M. A. (1976).Physiology of hibernating bears. In (M. R. Pelton,J. W. Lentfer & G. Edgar Folk, Eds) Bears—TheirBiology and Management, Third (1974) InternationalConference on Bear Research and Management,pp. 373–380. Morges: International Union forConservation of Nature, ICUN New Series, no. 40.

Gamble, C. (1986). The Palaeolithic Settlement ofEurope. Cambridge: Cambridge University Press.

Gargett, R. H. (1996). Cave Bears and Modern HumanOrigins: The Spatial Taphonomy of Pod Hradem Cave,Czech Republic. Lanham, MD: University Press ofAmerica.

Garshelis, D. L. & Pelton, M. R. (1980). Activity ofblack bears in the Great Smokey Mountains NationalPark. J. Mammal. 61, 8–19.

325

Glenn, L. P., Lentfer, J. W., Faro, J. B. & Miller, L. H.(1976). Reproductive biology of female brown bears(Ursus arctos), McNeil River, Alaska. In (M. R.Pelton, J. W. Lentfer & G. Edgar Folk, Eds)Bears—Their Biology and Management, Third (1974)International Conference on Bear Research and Manage-ment, pp. 381–390. Morges: International Union forConservation of Nature, ICUN New Series, no. 40.

Hellgren, E. C., Vaughan, M. R., Kirkpatrick, R. L. &Scanlon, P. R. (1990). Serial changes in metaboliccorrelates of hibernation in female black bears.J. Mammal. 71, 291–300.

Howell, F. C. & Arsebük, G. (1989). Report on the1988 investigations in the Cave of Yarimburgaz(Marmara, Turkey). Unpublished report submittedto the National Geographic Society, Washington,D.C.

Howell, F. C. & Arsebük, G. (1989/1990). Reporton the current status of research on the Cave ofYarimburgaz (Marmara, Turkey). Unpublishedreport submitted to the National Geographic Society,Washington, D.C.

Jenkinson, R. D. S. (1984). Creswell Crags, LatePleistocene Sites in the East Midlands. Oxford: BARBritish Series 122.

Johnson, K. G. & Pelton, M. R. (1980). Environmentalrelationships and the denning period of black bears inTennessee. J. Mammal. 61, 653–660.

Judd, S. L., Knight, R. R. & Blanchard, B. M. (1986).Denning grizzly bears in the Yellowstone NationalPark area. In (P. Zager, D. Garshelis, D. Graber, A.LeCount & C. Willey, Eds) Bears—Their Biology andManagement, Sixth (1983) International Conference onBear Research and Management, pp. 111–117.Washington, D.C.: Port City Press.

Koby, F. E. (1940). Les usures séniles des caninesd’Ursus spelaeus et al préhistoire. Verhandlungender Naturforschenden Gesellschaft in Basel, Vol. LI,pp. 76–95.

Koby, F. E. (1953). Modifications que les ours descavernes ont fait subir à leur habitat. Premier CongrèsInternational de Spéléologie, Tome IV(4), pp. 15–27.

Kuhn, S. L., Arsebük, G. & Howell, F. C. (1996).A Middle Pleistocene lithic assemblage fromYarimburgaz Cave, Turkey. Paléorient 22(1), 31–49.

Kurtén, B. (1958). Life and death of the Pleistocenecave bear: a study in paleoecology. Acta ZoologicaFennica 95, 4–59.

Kurtén, B. (1976). The Cave Bear Story. New York:Columbia University Press.

Lowe, V. P. W. (1967). Teeth as indicators of age withspecial reference to red deer (Cervus elaphus) ofknown age from Rhum. Journal of Zoology, London152, 137–153.

Lyman, R. L. (1987). On the analysis of vertebratemortality profiles: sample size, mortality type, andhunting pressure. Am. Antiq. 52, 125–142.

Lyman, R. L. (1994). Vertebrate Taphonomy.Cambridge: University of Cambridge Press.

Marks, S. A. & Erickson, A. W. (1966). Age deter-mination in the black bear. Journal of WildlifeManagement 30, 389–410.

Mattson, D. J., Knight, R. R. & Blanchard, B. M.(1992). Cannibalism and predation on black bears bygrizzly bears in the Yellowstone ecosystem, 1975–1990. J. Mammal. 73, 422–425.

McCullough, D. R. (1981). Population dynamics ofYellowstone grizzly bear. In (C. W. Fowler & T. D.Smith, Eds) Dynamics of Large Mammal Populations,pp. 173–196. New York: John Wiley & Sons.

Miller, S. D. (1985). An observation of inter- andintra-specific aggression involving brown bear, blackbear, and moose in southcentral Alaska. J. Mammal.66, 805–806.

Murie, O. J. (1951). The Elk of North America.Pennsylvania: Stackpole.

Özdogan, M. & Koyunlu, A. (1986). Yarimburgazmagarasi: 1986 yili Çalismalarinin ilk sonuçlari vebazi gözlemler. Arkeoloji ve Sanat (Istanbul) 32/33,4–14.

Peterson, R. O., Woolington, J. D. & Bailey, T. N.(1984). Wolves of the Kenai Peninsula, Alaska.Wildlife Monograph, no. 88.

Picton, H. D. & Knight, R. R. (1986). Using climatedata to predict grizzly bear litter size. In (P. Zager, D.Garshelis, D. Graber, A. LeCount & C. Willey, Eds)Bears—Their Biology and Management, Sixth (1983)International Conference on Bear Research andManagement, pp. 41–44. Washington, D.C.: PortCity Press.

Piperno, M. (1976–77). Analyse du sol Mousterien dela Grotte Guattari au Mont Circé. Quaternaria 19,71–92.

Poelker, R. J. & Hartwell, H. D. (1973). Black Bear ofWashington: Its Biology, Natural History and Relation-ship to Forest Regeneration. Washington State GameDepartment, Biological Bulletin no. 14.

Prat, F. (1988). Les Ursidés de la grotte Vaufrey. In(J.-P. Rigaud, Ed.) La Grotte Vaufrey: Paléoenviron-nement, Chronologie, Activités Humaines, pp. 291–318.C.N.R.S., Mémoires de la Société PréhistoriqueFrancaise, Tome XIX.

Reynolds, H. V., Curatolo, J. A. & Quimby, R. (1976).Denning ecology of grizzly bears in northeasternAlaska. In (M. R. Pelton, J. W. Lentfer & G. EdgarFolk, Eds) Bears—Their Biology and Management,Third (1974) International Conference on Bear Researchand Management, pp. 403–409. Morges: Inter-national Union for Conservation of Nature, ICUNNew Series, no. 40.

Rogers, L. L. (1981). A bear in its lair. Natural History90, 64–70.

Rogers, L. L. (1987). Effects of food supply and kinshipon social behavior, movements, and populationgrowth of black bears in northeastern Minnesota.Wildlife Monographs, no. 97.

Ross, P. I., Hornbeck, G. E. & Horejsi, B. L. (1988).Late denning black bears killed by grizzly bear.J. Mammal. 69, 818–820.

Spahni, J.-C. (1954). Les gisements à ursus spelaeus del’Autriche et leurs problèmes. Bulletin de la SociétéPréhistorique Française 7, 346–367.

Stiner, M. C. (1990). The use of mortality patternsin archaeological studies of hominid predatory

326 . .

adaptations. Journal of Anthropological Archaeology 9,305–351.

Stiner, M. C. (1991). A taphonomic perspective on theorigins of the faunal remains of Grotta Guattari(Latium, Italy). Curr. Anthrop. 32, 103–117.

Stiner, M. C. (1994). Honor among Thieves: A Zoo-archaeological Study of Neandertal Ecology. Princeton,NJ: Princeton University Press.

Stiner, M. C., Arsebük, G. & Howell, F. C. (1996).Cave bears and Paleolithic artefacts in YarimburgazCave, Turkey: dissecting a palimpsest. Geoarchaeol-ogy 11(4), 279–327.

Stiner, M. C., Achyuthan, H., Arsebük, G., Howell,F. C., Josephson, S., Juell, K., Pigati, J. & Quade, J.(1998). Reconstructing cave bear paleoecology fromskeletons: a cross-disciplinary study of MiddlePleistocene bears from Yarimburgaz Cave, Turkey.Paleobiology 24(1): 74–98.

Stoneberg, R. P. & Jonkel, C. J. (1966). Age determi-nation of black bears by cementum layers. Journal ofWildlife Management 30, 411–414.

Tassi, F. (1983). L’orso bruno Marsicano. Panda 12(Estratto, Parco Nazionale d’Abruzzo).

Tietje, W. D., Pelchat, B. O. & Ruff, R. L. (1986).Cannibalism of denned black bears. J. Mammal. 67,762–766.

Torres, T. J. (1988). Osos (Mammalia, Carnviora,Ursidae) del Pleistoceno de la Península Ibérica. Publi-caciones especiales de boletín geologica y minero(reimpresión del I-VI, T. XCIX, 1988).

Wainwright, S. A., Briggs, W. D., Currey, J. D. &Gosline, J. M. (1976). Mechanical Design in Organ-isms. Princeton, NJ: Princeton University Press.

Watts, P. D. & Jonkel, C. (1988). Energetic cost ofwinter dormancy in grizzly bear. Journal of WildlifeManagement 52, 654–656.

Watts, P. D., Øritsland, N. A. & Hurst, R. J. (1987).Standard metabolic rate of polar bears under simu-lated denning conditions. Physiological Zoology 60,687–691.

White, T. D. & Toth, N. (1991). The question of ritualcannibalism at Grotta Guattari. Curr. Anthrop. 32,118–124.

Wiszniowska, T. (1982). Carnivora. In (J. K.Kozłowski, Ed.) Excavation in the Bacho Kiro Cave(Bulgaria), Final Report, pp. 52–55. Warsaw:Pan’stwowe Wy Dawnictwo Naukowe.

Wolverton, S. (1996). Morphometry and Taphonomyof the Lawson Cave Ursids. Unpublished MastersThesis, Department of Anthropology, University ofMissouri-Columbia.