Mezmaiskaya Cave: A Neanderthal Occupation in the Northern Caucasus1

Cet article est consacre a la grotte de Mezmaiskaya, au nord du Caucase. Ce site occupe comporte une sequence stratifiee de niveaux du Paleolithique moyen et superieur que l'A analyse ici avant d'etudier les divers objets en pierre, os et ivoire ainsi que les restes d'un squelette humain et des ossements d'animaux decouverts dans la grotte.

Did Indo-European Languages other (earlier or later) population movements could Spread Before Farming? 1 have followed the same track.
While both the battle-axe and the farming-wave hypotheses have much to recommend them, they may j on atha n a d ams and m arc e l o tte not be the only reasonable explanations in terms of MS 6335,Environmental Sciences Division,Oak what is known of the prehistory of Europe and western Ridge National Laboratory, Oak Ridge, Tenn. 37831, Asia against a background of environmental changes. U. S.A. (Jonathan@elvis.esd.ornl.gov

)/Université de
The possibility that the initial dispersal event of the Liè ge, Service de Pré histoire, place du XX Aoû t 7, Indo-European languages involved not Neolithic farm-Bâ timent A1,4000 Liège,Belgium (prehist@ulg.ac.be). ers or Bronze Age warriors but Mesolithic hunter-11 vi 98 gatherers has been mentioned briefly by several writers (e.g., Renfrew 1987), but no one seems to have given the The question of how the Indo-European family of lanidea more than a passing thought. Here we aim to disguages came to occupy a broad swath of Europe and cuss this idea in greater depth, examining what is western Asia has long attracted discussion. The actual known of the climatic and the archaeological record, torange that the Indo-European family of languages had gether with general ecological principles of populations, achieved by early historical times is uncertain, but they to determine whether it stands up to more detailed were certainly present in central and northern Europe, analysis. southeastern Europe, Anatolia, and parts of the Near and Middle East. Celtic, Germanic, and Slavic migraclimate instability and language spread tions may have produced a relatively late overlay of Indo-European languages in parts of western and north-The past 100,000 years have been marked by many draern Europe, though without written records of the prematic climate oscillations (Van Andel and Tzedakis existing languages it is impossible to say what was 1996), each of which would have been capable of causwidespread before then. Migrations and conquest may ing changes in human population density as the relikewise have carried Sanskrit and Tocharian farther source base shifted. Episodes of relatively low populaeast shortly before early historical times. While action density-for example, during intense cold and dry knowledging that these identifiable movements of culphases-would have been followed by rebound periods tures and peoples contributed to the later spread of the in which humans could expand in range and in numbers Indo-European languages, scholars have long discussed across the region. As well as acting as a source of gewhat events before this time might have led to the netic shifts in population composition, the ''sampling group's present distribution. error'' caused by contraction of populations followed by Recent discussion of the prehistoric spread of the exponential expansion out of refuge areas could have Indo-European language group has generally concenproduced the spread of waves of linguistic and cultural trated on two sets of hypotheses. On one hand there is uniformity across the region. Just such a wave of poputhe view that migrations of warlike peoples (e.g., the lation out of the Near East may have carried Indo-Kurgan or Battle-Axe culture) (Childe 1950, Gimbutas European languages across much of Europe and some 1980) had spread the languages through conquest of reldistance into Central Asia. atively passive farming populations. A more recent al-Following a climate phase marked by low human ternative view (Renfrew 1987(Renfrew , 1992 is that the main population densities across the region, any one group event in the spread of the western branch of these lanthat acquired both the general cultural traits that guages was the initial spread of farming out of the Near caused it to spread rapidly out of a refugium and the East, providing a population ''wave'' (due to the intechnology to enable it to do so would have experienced creased carrying capacity of the farming lifeway) that rapid exponential population growth in an environment swamped the non-Indo-European languages of the relatively free of competition from other hunterhunter-gatherer groups that had previously existed in gatherer groups. Such a group, spreading out northward the area. This idea has received some support from geand westward and possibly eastward as well, would netic evidence of a southeast to northwest gradient in have made a disproportionate contribution to the genetic and linguistic legacy of Europe and parts of the Near East. Other groups even a few centuries slower in expanding in size and range in response to the climate 73 t abl e 1 change would have been numerically dominated by the earlier colonists as they left their refugial homelands, Sudden Climate Changes in Europe and the Near East during the Past 15,000 Years given the likely exponential growth rates of these populations. Even at the relatively low densities that huntergatherer populations would have been capable of Calibrated achieving, competition or at least interaction between Years Ago Climate groups would eventually have become more frequent, and the less numerous (non-Indo-European-speaking) 14,500 Warm and moist; rapid deglaciation groups would have been much more likely to lose their 13,500 At least as warm and moist as today's cultural and linguistic identity among a larger wave of 13,000 Cold (Older Dryas) Indo-European-speakers. This scenario may explain the 12,800 Ϯ 200 Transition to cold, dry (Younger Dryas) 11,500 Ϯ 200 Transition to warm and moist (Holocene, or existence of the Basque language group, which may isotope stage 1) have been a ''potential'' European dominant that nar-9,000 Warmer and often moister than today's rowly failed to expand before the Indo-European-speak-8,000 Slightly warmer and moister than today's ers became abundant in central and eastern Europe, 5,900 Cold (corresponding to ''elm decline '') southeastern Europe, and possibly also most other parts 4,500 Becoming fairly similar to present 2,600 Relatively wet/cold in many areas of Europe. The extinct Etruscan, Ligurian, and Iberian language groups may be regarded as further examples of the same. As the hypothetical Indo-European wave sources: Adams (1998), Adams, Maslin, and Thomas (1998). spread out in each direction, it can be expected to have gathered up the genetic and linguistic legacies of scattered smaller populations it encountered along its way Turkey the Younger Dryas period was even more arid than the most extreme part of the last glacial, with as each of these began a slightly later spread out of southern European refugia. This process of gathering up semidesert predominant. Conditions across most of the rest of Europe are variously thought to have resembled may explain some of the current east-west and northsouth genetic gradients which now exist in Europe and open cold forest steppe or possibly (at some stages) semidesert (Starkel 1991). A priori, such conditions may some of the differences between the present-day branches of the Indo-European family of languages.
be expected to have resulted in some change in human population densities and distribution, though it is dif-It is thus possible that much of the initial (mid-Holocene) range of the Indo-European languages across cen-ficult to demonstrate or disprove this idea given the limitations of the archaeological record for the Younger tral and northern Europe, the Balkans, and the Near East was achieved by the rapid spread of a sparse hunter-Dryas period. In Europe and most of the Near East the record of human occupation during the Younger Dryas gatherer wave out of either southern Europe, the Levant, Anatolia, or western Asia, preceding the farming is ambiguous, with the age plateau in radiocarbon ages adding to the confusion (10,000 years ago in radiocarbon wave. terms can mean anything between 11,200 and 12,200 ''real'' years ago). In northern and central Europe, the the potential importance of the younger record is perhaps detailed enough to suggest a complete dryas cold phase or almost complete depopulation during the Younger Dryas. However, in the Levant conditions seem to have An obvious candidate for an environmental change which could result in rapid and widespread change in remained relatively moist (Rossignol-Strick 1995), with relatively strong signs of continuity in human settle-languages, cultures, and genetic composition of human populations is the Younger Dryas cold event (about ment (the Natufian) (Henry 1989). Even in this area, for instance, in the Jordan Valley, aridity and a large de-10,800-10,000 14 C years ago), which returned much of western Asia to cold semidesert conditions (Huntley crease in food plants are accompanied by smaller human populations clustered around relatively moist and Birks 1983, Starkel 1991, Landmann, Reimer, and Kempe 1996, Jalut et al. 1988, Rossignol-Strick and ''oases'' (Wright 1993. Following the Younger Dryas, warm, moist Holocene conditions seem to have re-Planchais 1992, Velichko 1993), apparently through a series of rapid stepwise cooling events (see table 1). The turned rapidly all across Europe and western Asia, taking only a few decades according to the latest ice-core transition to the Holocene is marked by noticeable changes in technologies (to the Mesolithic) and in hu-indicators of regional climate (Taylor et al. 1997).
Given the magnitude of the change in environments man skeletal morphology across this region, possibly suggesting an immigration event. Reviewing the evi-across the region, the earlier climate transition (about 12,000 14 C or 14,500 calibrated years ago) to the much dence for hunter-gatherer carrying capacities in different environments, Steele, Adams, and Slukin (1998) colder, arid Younger Dryas could well have eliminated much of the previous Late Palaeolithic population of suggest that temperate forest and moist steppe have a much higher overall carrying capacity than either semi-northern and central Europe or at least drastically reduced interior population densities, and (from available desert or arctic environments. Drawing on a variety of sources, Rossignol-Strick indications of the carrying capacity of temperate forest environments for hunter-gatherers [Steele, Adams, and (1995) suggests that in many areas of Greece and across Volume 40, Number 1, February 1999 75 Slukin 1998]) the rapid return of warm conditions Thomas 1998) and to have come on (and also ended) over at most a few decades, lasting in total about 200 would have provided an opportunity for rapid human population expansion to fill this gap.
years. Here again, a decline in human population densities across much of the region seems plausible from the Renfrew (1987Renfrew ( , 1992 has vigorously attacked the techniques of linguistic dating and has found broad sup-extent and the suddenness of this event. Turnover in population or in cultural identity among hunter-gather-port among archaeologists if not among linguists. He makes the point that linguistic dating (based on degree ers resulting from this disruption might well have initiated or added to the spread of the Indo-European lan-of similarity in vocabulary and the use of specific ''technology'' words to pin down the culture of the earliest guages. Estimates of the linguistic chronology of the Indo-Indo-Europeans) is potentially subject to great imprecision. If one takes Renfrew's view that linguistic dating European languages have been used to suggest that much of their common vocabulary has a more recent of language history is unreliable, then an earlier divergence relating to hunter-gatherer recolonization after origin (about 7,000 years ago) (Swadesh 1972) than the early Holocene divergence that this sparse-hunter-gath-the Younger Dryas may be more plausible for a spread of Indo-European languages by this type of mechanism. erer-wave hypothesis (and Renfrew's farming-wave hypothesis) would seem to require (about 10,000-11,000 An error of about a factor of two in the estimated rate of divergence taken from the earliest written records years ago). In this sense, the more likely candidate is the later, less severe cold event 8,200 years ago. would be sufficient to push the point of common origin back several thousand years from the early-to-mid-There is also a possibility (though it conflicts still more strongly with the linguistic dating) that the popu-Holocene to the earliest Holocene. Given that during this time there was dramatic cultural change to rela-lation increase causing the initial spread of the Indo-European languages occurred at the earlier warming tively sedentary Neolithic farming (with lesser changes in trade patterns and technology) all across the region, event at the end of the Last Glacial Maximum (about 14,500 years ago), with the onset of the Younger Dryas one must ask whether the linguistic chronology is accurate. One can suggest, for instance (M. Fraser, personal itself, or perhaps at an even earlier event (Otte 1994).
One can also envisage scenarios combining aspects of communication), that relatively mobile hunter-gatherer populations moving across large areas of the European each of the three hypotheses. Quite independent of climate change, a more effective ''Mesolithic'' technology continent would have retained their cultural and linguistic unity more readily than denser and more seden-might have led directly to a population wave of increased carrying capacity analogous to that associated tary farming populations.
We do not claim that this particular hypothesis has with the Neolithic transition. As pointed out above, another major cold and arid event-lasting perhaps 200 any more merit than either the battle-axe hypothesis or the farming-wave hypothesis, but it should be seriously years-affected Europe and western Asia around 7,400 14 C 8,200 calibrated years ago (Alley et al. 1997) (table considered (given the uncertainty about the early linguistic history of the region) alongside these as another 1). An initial early Holocene sparse-hunter-gathererwave spread of the Indo-European languages might have possible scenario. Further light might be shed on this matter if the archaeological record of the region im-been followed by a period of relatively long-distance cultural and linguistic exchange (with possible spread of proved, allowing this hypothesis to be subjected to more rigorous testing. For example, good evidence of se-innovations in the language, continually updating aspects of the general substratum of Indo-European lan-vere depopulation of most of Europe and western Asia during all or part of the Younger Dryas or the early guages [sensu Sherratt 1996]) by relatively mobile hunter-gatherer groups and later farming and warrior Holocene cold phase would lend support to it, while lack of any depopulation might be seen as evidence groups.
A major refuge of population in the Europe-West against it.
Asian region during the Younger Dryas seems to have been the area of the Jordan Valley, where populations a plethora of population waves in the late clustered in moist sites in which wild nuts and grains glacial and holocene? could be gathered. It is interesting to consider that this region, having the general characteristics of a source re-The post-Younger Dryas colonization hypothesis is only one of a range of potential scenarios, suggested by gion for a sparse wave of hunter-gatherers, was also a key source for the farming wave. The Indo-European the paleoenvironmental and archaeological record, leading to the spread of Indo-European languages or of languages might thus have been propelled out of this source region by two successive population waves, first higher-order language groups such as the paired Indo-European/Uralic familes or the still broader and more one of hunter-gatherers and then a slower one of farmers. heterogeneous Nostratic superfamily. Another event that might have affected the spread of Indo-European by The idea that a phase of colder, drier conditions might have led to a regional decline in population den-either hunter-gatherers or early farmers or both is the widespread cold, dry event at 8,200 calibrated years ago sity depends crucially on the cultural habits and cultural flexibility of the hunter-gatherer inhabitants of (table 1). This event seems to have been about half as severe as the Younger Dryas (Adams, Maslin, and the region. One might hypothesize (as mentioned above) that an opening-up of the returning woodlands 8,200 or that of 5,900 calibrated years ago) would seem more consistent with observations. A change 8,200 due to cold and/or aridity would have favoured hunting of large animals, making possible an increase in popula-years ago could have promoted the spread of Indo-European languages by either hunter-gatherers or farm-tion density. This could merely push the dating of the necessary population wave slightly farther back in time. ers or both.
Alternatively, climate change may have had little or However, animal protein is not a principal source of food outside arctic and coastal environments, so it no role in the spread of the languages by farmers or post-Neolithic warriors. Different processes could coinci-would be unlikely to result in anything other than a net decrease in population elsewhere as the plant food dentally have aided the spread of the Indo-European language family at different times. It may be that an initial availability from wild grasses and nut-bearing trees declined with a shift towards cold steppe and semidesert sparse wave of recolonizing hunter-gatherers carried this group of languages part-way into central Europe conditions. The upshot of the complexity of human behaviours is that different cultural groups of humans and western Asia, with later processes such as the spread of farming and migrations of warrior cultures be-might have responded to the same change in opposite ways, and such a pattern of simultaneous decline of one ing responsible for its further spread.
Our hypothesis that climate change promoted the group and expansion of another might have produced a pattern of linguistic spread. In any case, overall popula-spread of Indo-European languages by causing changes in the population density of hunter-gatherer groups may tion density seems likely to have declined during cold, arid events, and the archaeological evidence from the be difficult or impossible to test. Given the a priori case which can be made, it seems fairly plausible, although Near East supports this.
Thus, if climate events were rapid and intense like Renfrew's farming-wave hypothesis it contradicts paleolinguistic analyses. The severity of this weakness enough to disrupt hunter-gatherer (and/or farming) populations, they may have been responsible for population is uncertain, however, as Renfrew (1987Renfrew ( , 1992 has pointed out various grounds on which the paleolinguis-or cultural replacements which helped to spread languages. Since the most intense events (the Younger tics can be doubted. The general hypothesis that past climate changes Dryas and the cold event of 8,200 years ago) precede the Neolithic across most of the Europe/Near East region, strongly affected linguistic patterns can also merge into more traditional explanations; sudden climate change hunter-gatherers may have been the vectors of the Indo-European languages.
could have been the primary cause of migrations of Indo-European-speaking Neolithic farmers or horse-This is merely speculation, but laying out possible scenarios is important to show that the situation in the riding warriors. If one accepts the paleolinguistic view that such ''technology'' words as ''wheel'' and ''copper'' region could well have been more complex than has generally been thought. The very variable environmen-were initially present at the point of divergence of Indo-European languages and that they applied to items such tal record of the Late Glacial and Holocene suggests that there would have been ample opportunities for as a fully formed wheel or worked copper, then the 8,200-or 5,900-year climate event (rather than the population and cultural/linguistic replacement quite unrelated to agriculture or migrations of warriors on Younger Dryas) could have been more important, respectively influencing migrations of farming groups or horseback.
of horse-riding warriors. The fact that one can so readily add and interchange alternative hypotheses concerning conclusion the spread of the Indo-European languages (and other language groups, all of which have formed in the highly The paleoenvironmental record suggests various times over the past 15,000 years at which major changes in variable world of the Late Quaternary period) should perhaps be seen as reason for scepticism regarding any hunter-gatherer population density could have occurred on a regional scale as a result of environmental changes. prospect of understanding the true nature of the initial spread of the Indo-European languages. Finding out Such population shifts would be difficult to detect in a sparse archaeological record subject to large 14 C anoma-what one does not know is, however, a vital part of the scientific process; it is always better to realise that there lies, but they remain a distinct possibility given the magnitude of the climate and ecological changes re-are grounds for uncertainty than to hold an unfounded belief that one knows the answer. This uncertainty is corded from across the region. While the ending of the Younger Dryas event seems particularly likely to have reason for open-mindedness as to the causes of the spread of Indo-European rather than any sharp division resulted in population waves in the approximate time range of the origin of the Indo-European languages, any into entrenched views. one of these prehistoric changes could have initiated the spread of the Indo-European language group (and in a broader sense the linked Indo-European/Finno-Ugric References Cited group). Given the existing dating and the detailed lin- ing to the lack of well-stratified and dated sites (Lyubin and Myshtulagty lagat (Weasel Cave) in the northsa l t z m a n, j. p. s e v e r i n g h a u s, m. s. t w i c k l e r, j. w. c. w h i t e, s. w h i t l o w, a n d g. a. z i e l i n sk i.
central Caucasus altered this situation (Golovanova, 1997. The Holocene-Younger Dryas transition recorded at Sum-Levkovskaya, and Baryshnikov 1990, Hidjrati et al. mit, Greenland. Science 278:825-27. 1997). In addition, new excavations at two previously v a n a n d e l, t. h. , a n d p. c. t z e k a k i s. 1996  skaya Cave and Monasheskaya Cave) yielded human maiskaya Cave in 1987. Excavations directed by the senior author were conducted during 1987-97 and have skeletal remains and other information about the Paleolithic inhabitants of the region (Belyaeva 1992, Lyubin exposed a total of over 40 m 2 . Several thousand artifacts and many thousands of well-preserved faunal remains 1994). The discovery and excavation of Mezmaiskaya Cave has contributed further to filling the gaps in our have been recovered from four Middle Paleolithic and three Upper Paleolithic occupation levels. Excavated knowledge about this remote but important area in Eastern Europe. sediment from occupation layers was water-sieved for recovery of small artifacts and bone fragments. In 1993 a partial human skeleton was found in the lowermost geographic setting and research history Middle Paleolithic layer, and in 1994 additional skeletal fragments were encountered in the uppermost Middle Mezmaiskaya Cave is situated in the northwestern Caucasus approximately 50 km south of the city of Mai-Paleolithic layer (Golovanova 1994, Golovanova andRomanova 1995). kop ( fig. 1). The cave is found at an elevation of 1,310 m above mean sea level at 44°10' N 40°00' E, in the Azish-Tau Range, which is part of the Lagonak Upland. stratigraphy and dating It overlooks the right bank of the Sukhoi Kurdzhips (a small tributary of the Kurdzhips River). Mezmaiskaya Mezmaiskaya Cave contains a deep succession of clay, loam, and rubble deposits that date to the Holocene, Cave is a solution cavern formed in Jurassic dolomite and is approximately 35 m in length, 8.5-9 m in height, Late Pleistocene, and possibly earlier ( fig. 2). The uppermost layer (layer 1) is a sandy loam with small weath-and up to 25 m in width (at the entrance, which faces southwest). ered rubble and contains a mixture of faunal remains and redeposited artifacts of Paleolithic and post-Paleo-Paleolithic remains were first discovered at Mez-  The Middle Paleolithic occupation levels appear to yielded an assemblage of typical early Upper Paleolithic artifacts and an AMS radiocarbon date (obtained on date to the Middle Pleniglacial (oxygen-isotope stage 3), although the lowermost remains could be older (possi-wood charcoal from a former hearth) of 32,000 years b.p. (table 1).
bly oxygen-isotope substages 5a-5d). Radiocarbon dates from layers 2, 2A, and 2B all fall within the later part Beneath these layers lie approximately 2 m of deposits containing Middle Paleolithic occupation levels and of oxygen-isotope stage 3, but these dates, obtained on bone and clearly lying at or beyond the effective range human skeletal remains. The upper Middle Paleolithic levels (layers 2 and 2A) are composed of clay and clay of radiocarbon dating, should be treated as minimum estimates. The date from the lowest Upper Paleolithic loam with small to large angular rocks and pebbles. In these layers, the predominance of alpine meadow forms level suggests that the Middle-to-Upper-Paleolithic transition at Mezmaiskaya Cave is probably synchro-among the rodent fauna and the high proportion of nonarboreal taxa among the pollen spectra indicate signifi-nous with the transition in other areas of Europe (Klein 1989, Mellars 1996. The uppermost Middle Paleolithic cantly cooler climatic conditions than at present. The lower Middle Paleolithic levels are made up of a silt-to-level at Myshtulagty lagat (Weasel Cave) in the northcentral Caucasus recently yielded a wood charcoal AMS clay loam with small to large angular rocks and numerous small pebbles (layer 2B) and a clay with occasional date of 36,000 years b.p. (L. R. Kimball, personal communication, 1998). small to large angular rocks (layer 3). The rodent fauna from these layers exhibits a more balanced mixture of The layers that underlie the Mousterian occupations have thus far failed to yield any evidence of human hab-forest and alpine forms, and climatic conditions appear to have been somewhat milder than during the deposi-itation. The uppermost unit (layer 4) contains a major rockfall horizon that may date to the Early Pleniglacial tion of layers 2 and 2A, although nonarboreal pollen is still predominant. All of the layers containing Middle cold peak (oxygen-isotope stage 4); alpine meadow taxa  . 3) to 40-50% in layers 2A and 2 zone at the base of this layer may reflect forest soil formation during the preceding Early Glacial (oxygen- (fig. 4). Canted sidescrapers are also numerous in all layers, but double and transverse forms are less common. isotope substage 5a?). The basal levels (layers 5-7), which are represented by clay loam and clays with vary-The majority of sidescrapers were prepared on flakes or bladelike flakes. Mousterian points, often with thinned ing quantities of rubble and pebbles, also appear to have accumulated under milder climatic conditions and backs ( fig. 3, 3), and endscrapers ( fig. 3, 10) are also present in all levels. Denticulates are rare. probably date to the earlier Late Pleistocene (oxygenisotope stage 5) or earlier. The consistently high levels The most significant contrast between assemblages from the lower and upper Middle Paleolithic layers is of potassium in these layers may reflect repeated forest soil formation (Baryshnikov, Hoffecker, and Burgess the high proportion of bifacial tools in the former. These include bifacial sidescrapers, which are plano-1996:316).
convex and typically have butts; only isolated examples have been found in the upper layers. The lower layers paleolithic artifact assemblages contain small handaxes ( fig. 3, 2) and bifacial foliate points, which are represented by two fragments in layer Over 5,000 artifacts have been recovered from the Paleolithic occupation layers in Mezmaiskaya Cave. 3 and a complete specimen in layer 2B ( fig. 3, 1). Limaces and ''Chokurcha triangles'' (bifacially worked These layers may be subdivided into three Upper Paleolithic levels (layers 1A, 1B, and 1C), and four Middle Pa-triangular tools found in the Crimean cave site of Chokurcha I) are also present. Endscrapers in the lower lay-leolithic levels (layers 2, 2A, 2B, and 3). Significant differences exist among the assemblages both within and ers exhibit partial bifacial working. Layers 3 and 2B also contain a large number of tools with converging edges, between the Middle and Upper Paleolithic.
The Upper Paleolithic artifacts, many of which were although most of these are unifacial or possess alternating retouch. excavated during 1997, have not yet been described in detail. The assemblages from layers 1A and 1B are The assemblages from the lower levels (layers 2B-3) are similar to those in many other Middle Paleolithic chiefly comprised of unretouched blades and flakes. Retouched lithic items include endscrapers, backed sites in Central and Eastern Europe. This industry is sometimes referred to as the ''East European Mico-blades, and Gravette points. Nonlithic tools include bone awls and bone and ivory points (ivory was proba-quian'' (Bosinski 1967, Gá bori 1976 or simply the ''East European Mousterian'' (Praslov 1984). Although bifacial bly imported from lower elevations). These assemblages are similar to those from previously reported Upper Pa-implements were recovered from Il'skaya I (Zamyatnin 1929), the assemblages from layers 2B-3 at Mez-leolithic sites in the Northern Caucasus that are believed to date to the final Upper Paleolithic (Amirkha-maiskaya Cave represent the first well-defined example of this industry in the Northern Caucasus. The assem-nov 1986). The assemblage from layer 1C is different and contains endscrapers on massive flakes and large blages from the upper levels (layers 2-2A) are comparable to those from several Middle Paleolithic sites in the blade fragments, dihedral and multifaceted burins, points (including points on microblades), and backed northwestern Caucasus (Monasheskaya Cave, Gubs Shelter No. 1, and Barakaevskaya Cave) which have blades. Nonlithic implements include bone awls and points. On the basis of both typology and radiocarbon been grouped together as the ''Gubs Mousterian Culture'' (Lyubin 1977(Lyubin , 1994. As at Mezmaiskaya Cave, dating, layer 1C appears to be an early Upper Paleolithic industry. Although the undated assemblage from Ka-these assemblages (especially at Monasheskaya Cave and Gubs Shelter No. 1) are thought to date relatively mennomostskaya Cave, which is located approximately 20 km northeast of Mezmaiskaya Cave and exhibits late in the sequence of local Middle Paleolithic industries. some similarities to the layer 1C assemblage, was tentatively assigned to the earlier Upper Paleolithic on typological grounds (Amirkhanov 1986), Mezmaiskaya Cave human skeletal remains has produced the first reliable evidence of an industry in this time-range in the Northern Caucasus.
Human remains recovered to date from Mezmaiskaya Cave include a partial skeleton from the lowermost A total of 3,645 lithic artifacts were recovered from the Middle Paleolithic occupation levels during 1987-Mousterian level (layer 3) and cranial fragments from the uppermost Mousterian level (layer 2). The partial 95 (table 2). The cores exhibit a pattern of parallel flaking on slightly convex surfaces. Cores were rejuvenated skeleton was found in quadrant M-26 near the cave entrance, where layer 1 lies unconformably over layer 3. by creation of a new striking platform and continued flaking on the same surface; two-sided cores are rare. A The skull was partly damaged, especially in the facial area, and displaced. The left scapula, humerus, and ra-preliminary analysis of striking platforms indicates that retouched platforms are uncommon, although the dius were preserved in anatomical sequence, as well as much of the vertebral column and ribs. The leg bones quantity of retouched platforms increases in the upper levels. True blades are very rare in the lower levels but were fully displaced. The body had been deposited on the right side, the head oriented to the north and the more common in the upper levels.
Tools from all levels reflect the predominance of lat-left arm extended along the trunk but flexed slightly at  the elbow; the position of the legs could not be deter-where it adjoins the lateral orbit. The occipital bone is differentiated from that of a modern human by greater mined. The skeleton occupied an area of 20 ϫ 40 cm; no traces of a burial pit were observed, and no artifacts dimensions, thickness, and density and by its rectangular form. The interior surface is only slightly concave or faunal debris were found in the immediate vicinity of the remains. However, small charcoal fragments (2-and has several small, deep fossae. The intrajugular process is absent, but the paramastoid process is distinctly 3 mm in diameter) were recovered from the sediment surrounding the skeleton (fig. 5).
expressed, and the sulcus for the sigmoid sinus is present although weakly expressed. The paramastoid pro-The age at death of the layer 3 skeleton was estimated on the basis of dental development. Fourteen dental cess is not round but extended in shape. The condyloid fossa is larger and more fully expressed than in a mod-crowns derived from deciduous teeth were recovered. Given that formation of the occlusal surface of the ern infant. The margin of the foramen magnum also exhibits differences from that of a modern infant; it varies crown of the second deciduous molar was incomplete and that the remaining teeth did not exhibit neck and in thickness and is slightly curved, apparently reflecting a foramen magnum of extended oval form (pear-shaped root development, the infant probably died between the age of a foetus of seven months and a newborn of two in a modern infant). A fragment of the right upper portion of the squama of the occipital was preserved and months.
Among the cranial remains, three bones (frontal, oc-may be differentiated from that of a modern infant by a significantly smaller degree of curvature. The sphenoid cipital, and sphenoid) have been described and compared with those of a modern human infant of similar bone is more massive than in a modern infant as a function of the greater size of several transverse dimensions age. The squama of the frontal bone is denser and the frontal eminence less pronounced than in a modern in-and other features. The anterior clinoid process is comparatively large and more medially placed than in a fant. The zygomatic process is substantially thicker modern specimen. The tuberculum sellae is rectangu-distal portion of the diaphysis exhibits a large transverse diameter that affects its shape, and the proximal lar, not partly circular, in form. The chiasmatic groove is deeper, and the carotid sulcus and lingula are mas-end of the diaphysis also exhibits differences from that of a modern infant. The radial tuberosity is well ex-sive. The greater wing displays a thicker zygomatic margin and the pterygoid process a broad base.
pressed and shifted medially and posteriorly because of the small curvature of the bone around the vertical axis. The postcranial bones also exhibit some significant differences from those of a modern infant. The distal The position of the nutrient foramen is shifted towards the center of the diaphysis. The ulna is somewhat segments of the limbs (radius, tibia) are relatively shorter than the upper segments (humerus, femur). Ra-broader at all points, but the most striking differences from that of a modern infant specimen lie in the struc-diographic analysis indicates that the medullary cavities of the limb bones are large. Three postcranial bones ture of the proximal diaphysis. The transverse and sagittal dimensions of the olecranon are greater, and the ir-(femur, radius, and ulna) have been described and compared with those of a modern infant of similar age. The regular trapezoid shape of the upper end differs from the subrectangular form in a modern specimen. The medial femur is large, and the curvature of the diaphysis is reduced owing to the increased transverse and sagittal di-surface of the proximal end is somewhat convex, in contrast to the concavity along the medial edge between mensions. The linea aspera is less pronounced and the distal end especially large. The intertrochanteric crest the coronoid process and the olecranon in a modern infant. The specimen exhibits a small degree of curvature. is comparatively small, reflecting the enlarged sagittal and transverse dimensions of the proximal diaphysis.
The cranial fragments from layer 2 were found in a small pit (40 ϫ 20 cm) located in quadrant N-19 that The radius is massive; the transverse dimensions are greater at all points, and the sagittal dimension is espe-was covered with a limestone block and extended into layer 2A and 2B to a total depth of up to 50 cm. It was cially great on the proximal portion of the bone. The not possible to determine if this pit had been excavated Neanderthals (Lyubin et al. 1986, Belyaeva 1992, Golovanova et al. 1995. Despite the proximity of this region by the inhabitants of the cave or had been formed by the impact of the limestone block. A total of 24 frag-to the Near East, where modern humans are associated with some Middle Paleolithic occupations, there is cur-ments were recovered from the pit. The fragments represent portions of the frontal and adjoining left and right rently no indication that modern humans manufactured any of the Middle Paleolithic industries of the parietal bones. The presence of an anterolateral fontanelle (sphenoid fontanelle) indicates an age at death of Northern Caucasus. Moreover, the recent redating of the Starosel'e child in the Crimea has eliminated any 1-2 years. The cranial fragments are characterized by a small degree of curvature, apparently due primarily to compelling evidence for Middle Paleolithic modern humans in Eastern Europe as a whole (Marks et al. 1997). 2 the effects of postdepositional deformation.
The partial skeleton from layer 3 may be classified as The remains from layer 3 are consistent with the hypothesis that the Neanderthals intentionally buried an archaic human with clear affinities to the Neanderthals of Western and Central Europe (Homo neander-their dead but typically without evidence for accompanying ritual. Although no traces of a burial pit were ob-thalensis or H. sapiens neanderthalensis). Among the traits exhibited by the cranial and postcranial remains, served around the human remains, the presence of a partial skeleton is unique in the Mezmaiskaya Cave de-the developed paramastoid process, oval shape of the foramen magnum, and short distal limb segments are especially significant (Trinkaus 1981;Rak, Kimbel, and 2. An isolated molar (left M 2 ) recovered from the Mousterian site Hovers 1994Hovers , 1996. Skeletal remains found in associaof Rozhok I (horizon 4), located on the Sea of Azov coast, lacks artion with other Middle Paleolithic occupations in the chaic features and may be derived from an anatomically modern Northern Caucasus (Barakaevskaya Cave, Monashehuman but according to A. A. Zubov cannot be firmly assigned to either modern or archaic Homo sapiens (Praslov 1968:83-84). skaya Cave, and Matuzka Cave) are also derived from posits. All other skeletal remains, including those of an-mals in the Middle Paleolithic occupation layers (table  3). Bison was especially common in the lower Middle imals that died of natural causes in the cave (e.g., cave bear), were dispersed and fragmented, although in rare Paleolithic levels, ranging from 52% to 76% of the total number of identified specimens in layers 3 and 2B. Less instances several bones were found in anatomical sequence (Baryshnikov, Hoffecker, and Burgess 1996). common taxa include red deer (Cervus elaphus), Kudaro cave bear (Ursus deningeri kudarensis), and marmot Much of the layer 3 skeleton was protected from the processes of dispersal and breakage, probably by inten-(Marmota paleocaucasica). Few data are currently available on the faunal remains from the Upper Paleo-tional burial; the absence of evidence for a burial pit may be the result of erosion, which has affected por-lithic levels.
In order to assess the degree of preservation, a sample tions of layer 3 in this part of the cave (i.e., near the entrance). Careful examination of the sediment sur-of large mammal bones (n ϭ 479) was classified according to the five weathering stages defined by Behr-rounding the skeleton failed to reveal any material evidence of ritual behavior that might have been associ-ensmeyer (1978). Ninety percent of the sample fell into either the ''stage 0'' or the ''stage 1'' category, indicating ated with the burial (e.g., grave goods, red ochre). a very low degree of weathering. Breakage was characterized according to categories established for long faunal remains bones by Shipman, Bosler, and Davis (1981). A sample of large mammal long bones (n ϭ 134) reflected predom-Many thousands of exceptionally well-preserved vertebrate remains have been recovered from all levels of inance of spiral, sawtooth, and V-shaped fractures, indicating that the majority of the bones were broken in a Mezmaiskaya Cave, and their analysis has yielded much information about the paleoecology of the site relatively fresh condition. Many of the bones from the Mousterian occupation levels exhibit traces of carni-( Baryshnikov andHoffecker 1994, Baryshnikov, Hoffecker, andBurgess 1996). Taxonomic classification of vore damage and stone-tool cut marks. Carnivore marks were observed on 5% of the bison bones and 10% of the remains recovered during the 1987-89 excavations revealed predominance of steppe bison (Bison priscus), goat and sheep bones; these percentages are low for carnivore accumulations (Baryshnikov, Hoffecker, and Caucasian goat (Capra caucasica), and Asiatic mouflon (Ovis orientalis) among medium-sized and large mam- Burgess 1996:328-29). Stone-tool cut marks were ob-tabl e 3
- served on 7% of the bison bones and 5% of the goat, Northern Caucasus. Lithic assemblages with bifacial tools have been documented in other parts of Eastern sheep, and red deer bones. The cut marks occur on a variety of axial and appendicular skeletal parts and reflect Europe, including the Crimea and the East European Plain, where they are sometimes referred to as the East-multiple stages of the butchering process, including dismemberment and filleting (Baryshnikov, Hoffecker, and ern Micoquian. Although bifacial tools have been recovered previously from Middle Paleolithic contexts in the Burgess 1996:330).
Measurement of dental crown heights on lower third Northern Caucasus (e.g., Il'skaya I [Zamyatnin 1929]), the assemblages from layers 3-2B of Mezmaiskaya molars provides an age (mortality) profile for bison (n ϭ 19) from layers 3 and 2B and a combined sample of goat Cave represent the first definitive evidence of such an industry in this part of Eastern Europe. Moreover, the and sheep (n ϭ 22) from all of the Middle Paleolithic occupation levels. These taxa yield a profile dominated stratigraphic position of these assemblages indicates that this industry occurs relatively early in the Middle by prime-age adults (or ''prime-dominated'' mortality profile [Stiner 1994]). Remains of juveniles (teeth and Paleolithic sequence of the Northern Caucasus (although absolute dating remains problematic) and dur-bones) are almost completely lacking among the bison sample. Such a profile is not consistent with the pattern ing an interval of mild climates.
2. A different Middle Paleolithic industry character-of scavenged remains, and these data, in conjunction with the tool cut marks, indicate that most of the bison, ized by a comparatively high proportion of blades, predominance of sidescrapers, scarcity of bifacial tools, and goat, and sheep remains represent animals hunted by the inhabitants of Mezmaiskaya Cave. Given the eleva-presence of some typical Upper Paleolithic tools overlies the bifacial industry (layers 2A-2). Both the strati-tion of the site, much if not all of this hunting activity seems likely to have occurred during the warmer graphic position and the accompanying absolute dates indicate that this industry is late in the Middle Paleo-months (a hypothesis supported by some seasonality data) (Baryshnikov, Hoffecker, and Burgess 1996).
lithic sequence of the Northern Caucasus and probably dates to the later part of oxygen-isotope stage 3. Similar lithic assemblages have been recovered from other sites conclusions in the region (classified as the Gubs culture [Lyubin 1977]), and these assemblages also appear to date to the Several important observations and conclusions can be reported from the excavation and analysis of remains late Middle Paleolithic. The significance of the contrast between the two industries at Mezmaiskaya Cave is un-from Mezmaiskaya Cave: 1. A Middle Paleolithic industry containing a high clear but may be related to climatic differences and associated changes in the use of the site. proportion of bifacial implements is present in the 6. An Upper Paleolithic assemblage (layer 1C) over- ---. 1994. Neandertal'tsy Gupsskogo ushchel'ya na Severnom Kavkaze. Maikop: Meoty. lies the youngest Middle Paleolithic level at Mezlyubin, v. p., p. u. autlev, a. a. zubov, g. p. maiskaya Cave. The assemblage contains large blades, mains of molten metal, anthropomorphic clay figurines, censers, portable ceramic altars, and ritual vessels make up the temple assemblage. v la d imi r io ne so v Imitative ritual in Sapalli-culture burials is seen in Samara Society for Culturological Studies, P.O. Box the use of anthropomorphic clay figurines, models of 3004, Russia. 5 vi 98 altars, sacrifice areas, methods of grave construction, burials of animals, and urn burials. The most distinc-Imitative ritual was very widespread in the mortuary tive of these features are the cenotaphs and the votive practice of Bronze and Early Iron Age societies in the bronze replicas of tools and weapons. process of social stratification. It is particularly striking More than 250 cenotaphs have been discovered in the in the Southern Uzbekistan complexes of the Protocemeteries of the Sapalli culture ( fig. 1, e). Typologically Bactrian Sapalli culture, 18th-10th century b.c. (Askathey can be divided into six main groups: (1) those with rov 1977; Askarov and Abdullaev 1983; Ionesov 1990, figurines made of wood or cloth; (2) those with clay 1992, 1993). This study is an attempt to determine the anthropomorphic figurines and other ritual objects; role of this ritual in the social life and mortuary practice (3) those with roof-shaped vessels; (4) those with immoof protostate society through the analysis of recent arlated sheep; (5) those with sheep and figurines; and chaeological data from the site of Djarkutan, the largest (6) those without grave goods. The social and ritual sigof the early Proto-Bactrian urban centers.
its cenotaphs contained traces of ritual fire (23.9%) and The Sapalli culture, being in a transitional phase of animal sacrifice (47.8%). its historical development, required a regulative-More than 300 bronze miniatures have been recovcompensatory mechanism for social adaptation. Its ered from approximately 150 Sapalli-culture burials at symbolic attributes can be viewed as filling the gaps in the sites of Sapallitepa, Djarkutan, and Bustan ( fig. 1, d ). cultural life created by the conflict between old and They include tools (e.g., knives, adzes, chisels, matnew power institutions in a period of transition. Drawtocks, shovels, ladles, sickles), weapons (e.g., daggers, ing on Ural-Altaic mythology, A. M. Sagalaev (1991: swords, points, celt-adzes), household articles (e.g., lad-132) suggests that ''archaic thought strives to compenders, spoons, plates), toilet articles (e.g., mirrors, razors), sate for reality at turning points: the vanishing material and unfinished or indeterminate votive artifacts. All object is replaced by its analogy.'' In the midst of epthese articles were made especially for burials and not ochal transformations, humans constantly extended for household use. They were made mainly by smithing ''the limits of reality through the observation of unrealof thin bronze wire. ity'' (Beylis 1983:29).
The regulation of ritual by the temple gradually de-A lengthy process of institutionalization of secular veloped into domination of the whole community by and sacred power preceded the spread of imitative ritual the centralized religious ideology. The features and disin the Sapalli culture and culminated in the establishposition of anthropomorphic clay figurines from Djarment of the palace and temple complexes at Djarkutan. kutan and Bustan cenotaphs appear to model mortuary Its Temple of Fire seems to reflect the incorporation of rites. For example, in cenotaph 12, Djarkutan 4B, two the novel imitative ritual subsystem into traditional such figurines were placed in front of a miniature clay practice. Almost all of the various evidences of imitaaltar precisely replicating the round altars of the temtion in burials (votive replicas, cenotaphs, sacrifices, ple. The cenotaph also contained a miniature clay vesetc.) are represented in the temple. Moreover, the temsel with a tiny spoon and several cone-shaped clay obple itself was the embodiment of the imitative-ritual jects ( fig. 1, b). Apparently prayer by mourners was system; the complex and multileveled world of relibeing represented here. gious symbols was clearly reflected in its architecture, Votive miniatures are found in the cult centers of plan ( fig. 1, i), and contents (Askarov and Shirinov other societies. For example, bronze replicas of swords, 1991). It seems to have regulated in detail the sacred life arrowheads, and vessels have been discovered at conof the community. Its plan includes a treasury or relitemporary Meligeli 1 in eastern Georgia (Pitshelauri quary, a storehouse for sacred ashes, wineries, a corral 1979:52) and at Shilda, dating to the 14th-13th century b.c. (Pitshelauri 1982:58-59). A collection of votive Hissar in northern Iran. The emergence of such symbols stratum of the community. For example, the innovative types of tools and weapons, the novel forms and orna-in mortuary rites is connected, according to Dyson (1986:91), with the increasing differentiation of the so-mentation of the vessels, and the practice of disarticulating corpses were characteristic of the graves of the ciety. Bronze votive objects were very common in the Luristan archaeological complex, dating to the end of prosperous. Moreover, with the Sapalli culture for the first time bronze replicas of axes, adzes, chisels, and lad-the 2d and the early 1st millennium b.c. (Dandamaev and Lukonin 1980:61-67). The materials from the Ro-ders appear in three rich burials (graves 85 and 89 at Sapallitepa and 31 at Djarkutan 5). The emergence of the man cemetery at Osteria dell'Osa, with its votive bronze tools and weapons, miniature anthropomorphic Temple of Fire greatly expanded the sphere of imitative ritual activity, for the temple became a powerful accu-figurines of clay, ritual vessels, and other objects, demonstrate clear analogies with the imitative articles at mulator of religious ideas. The monumental temple and palace, bronze votive replicas, cult vessels, anthropo-Djarkutan (see Bietti Sestieri 1994).
To all appearances, the Djarkutan religious elite used morphic clay figurines, and models of altars testify to the development of an elite culture within Djarkutan imitative symbols in mortuary rites as a way of representing the class/prestige of the deceased. Indeed, socio-society.
Interpreting the mythological content of imitative cultural change in Sapalli society began with the upper ritual is a complex issue. Using the Sapalli culture as Cenotaphs including such vessels may exceptionally have contained the ashes of the deceased. an example, closer attention might be paid to a further very important aspect of the mythological semantics of It is obvious that the votive objects were magical things; through them people tried to influence their im-mortuary rites-the development of beliefs about the afterlife. Here the imitative rites stressed the spheres of mediate environment, and ''on being included in ritual they began to play an active social role'' (Iordanskij 1982: life and death in the mythological-ritual area representing the world. This intention is clearly evident in Djar-46). The increase in the social importance of the means of production in the technological process gradually gave kutan's historical topography and the development of its cemeteries. In the period just before the construction magical power to tools. The data from Sapallitepa, Djarkutan, and Bustan, however, confirm Propp's (1986:192) of the temple, the cemeteries were located on a mound outside the settlement, separated from it by a river that thesis that the earliest such magical objects were animal parts, with the part standing for the whole and the idea is interpreted as the personification of the sacred element of water, the mythical boundary between life and replacing the image. In the first stage of the Sapalli culture the human corpse was imitated by the body of a lamb or death.
Ideas of rebirth after death were originally reflected in sheep in the majority of cenotaphs ( fig. 1, e). At this point the animal may have played the dual role of double of the the Sapalli culture by interments of disarticulated bones. At least ten burials of this kind have been recog-deceased and animal helpmate. Afterwards, there were dolls and anthromorphic figurines and then votive repli-nized in the Djarkutan and Bustan cemeteries. The earliest such burial, dated to the late Kuzali phase, ca. cas of clay altars, ritual vessels, bronze tools, etc.; in short, one imitation led to another. 1250-1200 b.c., is Djarkutan 4B grave 7 ( fig. 1, h). In these burials the human bones are carefully arranged on The development of imitative ritual was determined primarily by the emergence of a new system of values. the floor of the grave and accompanied by a large number of ceramic vessels and bronze replicas of tools and Through social transformation a special system of symbolic communication was created in the mythological-weapons. The skeleton was usually sprinkled with red ochre. Male skulls are laid on their left side and female ritual area. This process reflected the historical conditions of a transitional stage in the development of skulls on their right. These burials contrast sharply with traditional Sapalli-culture burials, reflecting an in-ancient civilization. tentional violation of the usual rules. Disarticulation of the skeleton is an element of ancient myths from a wide References Cited variety of cultures. It is usually deities and heroes-for example, Buddha, Adonis, Osiris, Zagrey, Dionysos, a s k a r o v, a. 1977. Drevnezemledel' cheskaja kultura epokhi Pentheus, Orpheus, and Jima-that are cut into pieces one, dozens of stone beads, 2 miniature marble bowls, b e y l i s, v. a. 1983. ''Teorija rituala v trudakh Viktora Tur-2 baskets, 19 ceramic vessels, and a number of animal nera,'' in V. Turner: Simvol i ritual, pp. 7-31. Moscow: Nauka. d a n d a m a e v, m. a. , a n d v. g. l u k o n i n. 1980. Kultura i bones were found in female burial 82 at Sapallitepa. It Cylindrical vessels with roof-shaped lids were discov----. 1993. ''Imitatsija i obrjad,'' in Uzbekistan v drevnosti i ered in some burials at Djarkutan and Bustan ( fig. 1, c). tan 6 is particularly significant (Avanessova 1995).