I am thanking Dr. Larry D. Agenbroad , Emeritus Professor of Geology at the University of Flagstaff in Arizona and director of the Mammoth Site, as well as the entire team of the museum, for inviting me to study this exceptional collection of fossil of Proboscidea and for welcoming me so warmly, on the site as in their research laboratory. A big thank you to Gary Haynes for his kindness and valuable data. Finally, thank you to William Devriendt (cad- dap Douai) for this new illustration, here the reconstitution of Hot Springs (fig. 2).
1Discovered by chance in 1974 during the construction work of a housing scheme, the paleontological site of Hot Springs in South Dakota was completely preserved. If other sites have revealed remains of mammoths in the region, such as the Upper Palaeolithic site of Lange Fergusson also known worldwide (two butchered mammoths associated with Clovis points within an ancient swampland, Hannus 1989), the Site of Hot Springs is undoubtedly the most impressive judging by the concentration of skeletons of mammoths found in situ (fig. 1). Indeed, the site shows a significant accumulation of remains belonging to a typical upper Pleistocene fauna, largely dominated by the Columbian mammoths that were roaming the Great Plains of North America about ± 26,000 BP (average 14C dates, Laury 1990).
2The lake fed by a hot spring and the green bordering vegetation, typical of a swamp, would have been especially attractive during the glacial periods for animals and especially for mammoths whose daily needs must have been around at least 200 litres of water. The spring would have been a formidable trap for animals that ventured into this swamp with steep and slippery banks ending as a swallow hole in its submerged part. Thus, many animals were caught into the trap that proved fatal to the megaherbivores such as mammoths.
3Knowing the origin of an elephantidae bone assemblage is far from easy. Thus, the use of a solid frame of reference in terms of natural mortality, excluding the anthropogenic factor, is essential for the analysis of remains from paleontological sites. This requires knowledge of the social structures of present-day populations of elephants as well as of the mortality profiles issued from fossil species from purely paleontological sites. The aim of this study is to examine the possible circumstances of the death of the mammoths on the site given the analysis of the newly established mortality profile.
Figure 1 - Location of the main Mammuthus columbi sites.
4The site is located south of the Black Hills (GPS 43.4291, -103.494774). It was created out of the collapse ± 26,000 years BP ago (Laury 1990) of the ceiling of an underground cave (fig. 2). The 40 m long elliptical depression thus formed was presenting a wall whose inclination ranged from 60o to the horizontal and a depth varying from 14 m for the south-western portion to 21 m in the north-eastern part. There, a hot artesian spring supplied over time a lake whose surface temperature was around 35 to 40°C, as proven by malacological, pollen and fish analyses (Agenbroad 1990).
5The stratigraphic filling of the bowl, born from the collapse of the karst, is a superposition of laminated fine-grained sediments layers, ranging from layers of clay to coarse sand layers resulting from the erosion and weathering of the rock face and of the gravel making the ancient terrace of the neighbouring palaeoriver. The skeletal remains were unearthed from the three major phases of sedimentation of the sequence, and even if the excavation is still in progress, the main areas have already been tested in depth. The trap would have operated for about 300 to 700 years before its total filling by the sediments (Laury 1990).
6Swaying rotting carcasses and the movements of underwater gravity are factors not to be overlooked in the process of deposition of the bone remains. Indeed, they have actively contributed to the disarticulation and specific location of some specimens. Whatever the sedimentation stage considered, the most notable density of remains is located along the ancient banks. The bones, all recorded in three dimensions, are concentrated less than 6 m away from the wall (Laury 1990, fig. 2). It is possible that the artesian flow contributed to deprive the central part of the filling of remains. The north-eastern portion of the deposit has a high concentration of isolated bone remains from a sandy matrix, a consequence of the artesian flow, whereas the south-western part presents elements in anatomical connection set in a context of laminated layers of clay and silt, suggesting a much more moderate underwater energy. Thus, most of the carcasses were partially disarticulated before being dragged down to the bottom of the sunken bank and afterwards buried under sediments. The most remarkable specimen unearthed is the complete post- cranial skeleton of an adult female preserved in situ in anatomical connection (no 13S23W).
7The gradual erosion of the banks has allowed the excellent preservation of all the bones. Even the smallest and most delicate, such as the hyoid bone or sesamoid bones were preserved by the rapid covering by the sediments of the anatomical elements just discomposed. In addition, the natural origin of the fracturing of fresh bone, without impact or repercussion, highlighted on eleven Elephantidae bones, leaves little doubt. This is corroborated by the absence of anthropogenic traces on the site, as Agenbroad LD (1989) already concluded. It is possible that the still fresh bones broke under the weight of the mammoths in difficulty on the banks.
8While the overwhelming overrepresentation of Columbian mammoths (Mammuthus columbi) tends to obscure the diversity of the fauna documented at Hot Springs, many animal species belonging to mammals, birds, fish, amphibians and invertebrates have already been inventoried (Agenbroad, Mead, 994, updated on http://www.mammothsite.com).
9Among the other large mammals present at the site, the following were uncovered: woolly mammoths (Mammuthus primigenius), ancient camels (Camelops hesternus), llamas (Hemiauchenia sp.), pronghorn antelopes (Antilocapra americana) and finally five carnivores: giant short-faced bears (Arctodus simus), wolves (Canis lupus), coyotes (Canis latrans), badgers (Taxidea taxus) and another smaller mustelid mink or ferret-like (Mustela sp.). Two species of lagomorphs have also been found, namely Townsend hare (Lepus cf. Townsendii) and a rabbit (Sylvilagus sp.). Insectivores are represented by the mole (Scalopus aquaticus), while a number of rodents is consistently enriching the faunal list. These include first of all the prairie dogs (Cynomys ( Leucocrossuromys ) sp.), thirteen-lined ground squirrels (Spermophilus cf. tridecemlineatus) and slightly larger squirrels (Spermophilus cf. elegans or richardsonii), followed by the northern pocket gopher (Thomomys talpoides), hamsters (Neotoma cf. cinerea), mice (Peromyscus cf. maniculatus), heather voles (Phenacomys cf. intermedius) and smaller voles (Microtus sp.).
10If it seems remarkable to have unearthed Mammuthus primigenius (represented by three individuals) and Mammuthus columbi on the same site, this does not necessarily imply that the two species coexisted. Indeed, the skeletal remains of woolly mammoths come from the upper sediments. This could indicate that they have come after the Columbian mammoths at the site (Agenbroad, Mead 1994).
Figure 2 - Hot Springs restitution, ± 26,000 years BP (© W. Devriendt).
11The pond, fed by the hot spring, was located near the southern boundary of the North American ice sheet. Its bordering lush vegetation, typical of a swamp, would have been very attractive to wildlife. The surroundings were occupied by large areas of low vegetation, steppe-type, characteristic of the relatively cold general climate, to the detriment of the forest that only persisted residually at the bottom of the valleys.
12Ecological methods, based on the presence of mammals, were applied in order to come closer to the reality of the environment of the site at the time. These methods, ecological histograms and cenograms, have the advantage of not taking into account the quantification of the species but only their presence. The use of these tools is therefore particularly suitable for Hot Springs, which is still being excavated and whose complex sediment dynamics does not preclude a distribution of the remains by taxonomic format.
13The method of compiling and interpreting different ecological histograms takes into account the systematic of each species, but also the mass, the diet and locomotive adaptation of each of them. Defined by T.H. Fleming on the Central American and North American (Fleming 1973) current mammalian communities, the method was then transposed to fossil species by P. Andrews et al. (1979) and modified by Cl. Guérin and M. Faure (1987).
14The predominance of rodents in Hot Springs, followed to a lesser extent by carnivores and artiodactyls, characterizes an environment dominated by forest. However, the hegemony of the herbivores with hypsodont teeth indicates the openness of the environment (tabl. 1, fig. 3 and 4a and b). This diversity of vegetation cover is finally strengthened by the importance of species under a kilo, followed to a lesser extent by the under 10 kg, coupled with the notable presence of species over 200 kg (Fig. 4c).
Table 1 - Legend of ecologist histograms from Hot Springs.
Figure 3 - Ideal ecological profiles, according to Péan 2001 (cf. légend table 1)
Figure 4 - Ecological profiles from Hot Springs (cf. légend table 1).
15The study carried out by S. Legendre on current faunas showed a strong correlation between the weight distribution of mammals and their environment. The cenogram method of J.A. Valverde (1964), as amended by S. Legendre, was to see graphically on the abscissa the different taxa , excluding carnivores, ranked in descending order of weight, and on the ordinate the logarithm of their mass in grams. The estimate of the average mass of most fossil species was done by S. Legendre according to the surface of the first lower molar according to a specific formula and to constants specific to the taxonomic groups (Legendre 1989). On the approximate weight of certain species other references were considered (table 2).
16While species between 500 g and 5 kg are abundant in closed forest environment, they are absent or exceptional in the steppe open environment, resulting in a break of slope on the cenogram (fig. 5). Three taxa among the seventeen, excluding carnivores, identified at Hot Springs fall into this weight range, namely hares, rabbits and prairie dogs (appendix 2, fig. 6). Their presence suggests the existence of a mixed environment in which forest is not far behind. In addition, the more humid the context the more the ecological diversity is favoured, resulting in a smoother slope than in the case of a semi- desert environment (Legendre 1989). The slope is irregular in our case and the list of mammalian species rather limited so far. However, excessive aridity would likely have limited the presence of the very large species, which is not the case here.
Table 2 - Legend of cenogram from Hot Springs.
17Ecological histograms and cenograms from Hot Springs tend to characterize an immediate mixed environment mixing steppe and forest under a generally sub-arid climate. Our results are consistent with all the palaeoenvironmental data that characterize a green island in the heart of the steppe (Mead et al. 1990). The pollen analysis thus indicates that 46% of pollens come from grasses (Poacea), although wetland plants of the sedge type (Carex), aquatic plants (Potamogeton, Polygonum), aromatic plants (Artemisia) or flowering plants (Ambrosia, Chenopodiaceae) are also found. As for the tree species, less abundant, the pollen analysis indicates the presence of spruce, followed by broad-leaved trees (birch, juniper, walnut, elm, oak).
Figure 5 - « Cénogrammes » interpretation, according to Montuire, Marcolini 2002.
Figure 6 - « Cénogramme » from Hot Springs.
18Among the skeletons of Mammuthus columbi with anatomical connections, the most complete is undoubtedly the one that has the number HS00140, its head having remained in close connection with the rest of the body (fig.7d). The 13S23W specimen, assigned by A. Lister to a female, has a near-complete in situ post -cranial skeleton. The 87-016 skull could be associated with it given its proximity and the distance between its occipital condyles (Lister in Agenbroad, Mead 1994).
Figure 7 - Mammuthus columbi cranial vestiges from Hot Springs.
19Among the twenty-three skulls studied by L.D. Agenbroad in 1994, nine did not allow the attribution of age, against one mandible out of the twelve taken in consideration at the time (Agenbroad, Mead 1994). Among the total of forty-eight jaws at the end of 2005, seven did not allow us access to the molars. Only four skulls out of the twenty-four examined and three mandibles out of twenty-four once again did not give us access to the teeth (tab. 3 and 4 - Appendix). As many upper jaws as mandibles were thus taken into account without one knowing if it was possible or not to associate them. One hundred and eleven cheek teeth analyzed thus belong to dental series (two consecutive teeth, exceptionally three) against forty-six isolated molars, or a total of one hundred and fifty seven cheek teeth. Since late 2005, several postcranial elements have been exhumed. However, apart from the tusks, only one additional skull was unearthed during the excavation in 2009 (http://www.mammothsite.com).
20Two difficult-to-access mandibles as well as twenty-five isolated dental remains are of indeterminate age. They are in thirteen cases fragments of molars consisting in a few lamellas and in six cases of single lamella. In the case of four whole molars partially concealed by the sediment, the uncertainty of the tooth row within the cheek teeth series can only give a possibility of origin (for example, M2 or M3). A tooth at the end of its life is finally so abraded than its origin is also questionable. Among the whole molars, it can be noted that one is around six to fifteen years old while the others are from individuals over twenty years old. We will put in perspective the importance to give to these undetermined isolated teeth. Indeed, it is likely that they are from jaws for which the age group was offered. Out of one hundred and fifty seven cheek teeth examined, in place or not in the jaw, no deciduous tooth is attested, however. The teeth are exclusively M1, M2 and M3 molars, and especially the last and penultimate molars (tab. 3 - in appendix).
21Let us recall that in Elephantidae dental renewal is quite special. Mammoth, like the elephant today, had in total twenty-four cheek teeth. Thus, per half-jaw, six cheek teeth succeeded each other during the life of the animal, that is to say three deciduals (D2, D3, D4 also called m1, m2 and m3 by some authors) and three molars (M1, M2, M3). The replacement mode of the cheek teeth is quite original in Elephantidae because it happens in an “horizontal" manner actually following a very specific curvature. Thus, when a cheek tooth is substantially worn down by an abrasion phenomenon, it is pushed by the following and gradually disappears in favour of the latter, larger and at the beginning of its wear. Thus, there are never more than two functional teeth at a time on a half - jaw, maxillary or half-mandible. This sequence only comes to an end when the last permanent molar has also finished wearing down. Then the animals, unable to feed themselves, starve, approximately around 60 years old in current species (Laws 1966).
22The work to which we mostly referred to determine the position of Mammuthus columbi cheek teeth in the dental series are those of H.F. Osborn (1942), J.J. Saunders (1970) and C.T. Madden (1981).
23As the biometrics of the molar vary according to the degree of wear, it is imperative to consider the latter in order to put length, width or height measurements into perspective. In our study only measurements taken at a stage of wear allowing to keep the maximum dimensions of the tooth were retained (ideally wear stages A to C, fig. 8). By length (Li) of the tooth, we mean the length taken at the base of the tooth crown (Beden 1979). Other authors, such as V.J. Maglio (1973) or J.J. Saunders (1970) chose to measure the length (L) at mid-height of the tooth and perpendicular to the lamellas, and others parallel to the abrasion surface (in Agenbroad, Mead 1994 - p. 183). The synthesis of the measurements obtained on the molars of Mammuthus columbi at Hot Springs is shown in table 5.
24We can note that the position of some of the jaws in the sediment has not allowed taking measurements of the molars. This is the case of four skulls and three mandibles of the “eastern part” and the “south-eastern part”, all from adult individuals remained in situ on site (tabl. 4 - appendix).
Table 5 - Synthesis of measurements obtained from cheek teeth of Mammuthus columbi, Hot Springs.
25A molar, even if it is not whole, provides a number of very useful information, such as the species to which the animal belongs but also the age at death. Thus, the study of tooth wear of each cheek tooth tells us about the age at death of the individuals. This analysis is therefore the preliminary step to the establishment of mortality curves.
26Elephantidae cheek teeth consist of a series of more or less large transverse lamellas held together by the cement. These lamellas are themselves formed by a thin fold of ivory covered with a thicker enamel layer. The lamellas wear down nearly perpendicularly during abrasion, thus their size decreases progressively, leading therefore to reduce the height of the tooth. Several methods were proposed in order to assess the degree of wear of the Elephantidae cheek teeth. Among them, we will mention the Nf / N ratio frequently used (number of lamellas opened / total number of lamellas). However, if this formula is more than convincing regarding the whole molars, it is not applicable to dental pieces.
27Knowledge of the tooth wear stage is a very important parameter because the only position of the molar in the cheek teeth series is not sufficient to assign an age to the animal, or else it would be too approximate, especially for the last molars.
28We have opted for R. M. Laws’ classification of wear stages (1966) obtained on current elephants, coupled with the method for which M. Beden (1979) created a theoretical model applicable to “definitive” lower molars and we similarly established the one applicable to upper molars (fig. 8). To do this, almost two hundred mammoth upper cheek teeth from the site of Hanhoffen in the Bas-Rhin, Germany, were thoroughly analyzed (Louguet-Lefebvre 2005, 2006). The stages of wear thus defined by M. Beden range from new tooth (stage A) to end of life tooth, abraded at 100% (stage D4), passing through the stages of progressive destruction of the lamellas (stages D1 to D3 included). This method, like that of R.M. Laws, takes into account both the relative amount of lamellas in use and the estimation of missing lamellas due to wear, but it also takes into account the angle of the abrasion table with the dental crown. This last parameter is especially useful in the case of incomplete molars. That is why the correlation between this method and R.M. Laws’ is interesting. It allows assigning a relatively precise age group to tooth fragments, including upper ones (table 6).
Table 6 - Attribution of main age groups to Elephantids.
29The small existing differences between the ages assigned by different authors may impact the mortality profile, which is why it is crucial to note to which methodology we refer to. For example, the
XVl class defined by R.M. Laws was attributed to an age close to 26 years old, thereafter G.C. Craig matched it an approximate age of 19 to 21 years old (Craig in Haynes 1991). This is perhaps what partly explains the differences observed between the mortality profiles of Hot Springs established by different authors, even if the constant growth of the studied sample remains an undeniable profile modification factor. The estimation of age that we propose here is based on the model established by R.M. Laws on African elephants (1966), as current elephant have a dental formula similar to that of the mammoth and a life expectancy supposedly close.
30The most complete specimen (HS00140) had the advantage of having its two jaws. The analysis of the mandible and of the maxilla enabled us to assign to the individual an approximate age that should be around 40 to 45 years. Its last lower molars have nearly three quarters of open enamel lamellas and no lamella disappeared through wear, that is to say a B4 stage wear corresponding to the morphological stage XXIII defined by R. M. Laws.
31Several authors have previously studied the molars of Mammuthus columbi from Hot Springs in order to develop a mortality profile at a T instant of the excavation. Thus B. Dutrow in 1980 studied the forty-six cheek teeth excavated at that time and determined a MNI (minimal number of individuals) of eight (Dutrow 1980). In 1987, L.D. Agenbroad and J. Mead determined from the molars no less than twenty- six individuals (Agenbroad, Mead 1987) and forty-three from the count of the tusks carried out in 1990, and that for only twelve humerus (Agenbroad 1990). The study of the eighty molars then allowed L.D. Agenbroad to conduct the revision of nearly half of the ages that were previously determined by B. Dutrow and to establish a new mortality profile, while the MNI was then estimated to fifty-nine individuals (Dutrow in Agenbroad, Mead 1994). However, this is a frequency MNI and not the combination MNI we are referring to. Insofar as isolated teeth are considered, resorting only to MNI frequency has the disadvantage of significantly overestimating the number of individuals less than 40 years old. Indeed, these individuals have two teeth in use on each half- jaw against a single last molar beyond this age (M3). If we consider for example two isolated upper left teeth, more precisely an M1 with a wear stage C and M2 having reached the A wear stage, it is possible that these two teeth are issued from the same individual. The resulting profile then presented a clear overrepresentation of individuals aged 10-19 years and 20-29 years, as these age groups concentrated 80 % of the data, the remaining 20% being divided almost equally between the last three age groups. All authors agree in any case on the fact that no mammoth young was unearthed at Hot Springs, as evidenced by the absence of deciduous molars. Judging by the count of the tusks made to date by the L.D. Agenbroad foundation, the number of Columbian mammoths could reach fifty seven individuals in 2012 (http://www.mammothsite.com).
Figure 8 - Attrition levels of Elephantid cheek teeth, according to M. Beden, 1979 (lower teeth) and S. Louguet-Lefebvre, 2005 (upper teeth).
32The ages allocated to ten skulls by A. Lister by using the measurements provided by the occipital condyles (Lister in Agenbroad, Mead 1994) corroborate the ages we obtained on these skulls by studying their cheek teeth. Moreover, our analysis allows clarifying the results from the osteological studies carried out by L.D. Agenbroad (1990) which indicated an overrepresentation of males aged between 10 and 35 years.
33To determine the MNI (minimal number of individuals), we obviously took into consideration the possibility that an upper jaw (or two maxillaries), a mandible (or two hemi- mandibles) and one or more isolated teeth could come from a single individual. To date, only the HS00140 specimen has both jaws, also clearly associated by anatomical connection with the rest of the post -cranial skeleton, although other maxillary - mandible associations are possible. To determine the MNI we first added the maximum number of teeth of the same type (M3, M2), in strictly identical positions (upper or lower, left or right) and with an almost identical degree of wear by accepting a margin of evaluation in a stage (i.e. B4 for a tooth and C for another). We then divided our MNI by age category referring to table 6. The combination MNI then consisted of the addition of the respective MNI of each age category, that is to say thirty-three to which we added three, as three non accessible mandibles because of their position in the sediment come from individuals of unknown age (table 7). This calculation allowed us to review the latest estimates downward, since the cheek teeth come from a minimum of thirty- six individuals. This minimal number is without common measure with that provided up to date by the tusks, which could reach fifty seven. The excavations in the coming decades should help to gradually diminish this gap.
34Our review of the mortality profile of Mammuthus columbi, in light of recent data, indicates that the trap has affected a majority of adult subjects, since twenty-two individuals out of a total of thirty-six identified by their molars died at an approximate age between 25 and 38 years (table 7, fig . 9). The absence of deciduous teeth, as well as the lack of M1, not or slightly worn, confirms the absence of very young individuals and juveniles under the age of 10. The deficit in senile individuals associated with the lack of very young mammoths is surprising.
Table 7 - Mammuthus columbi from Hot Springs in MNIc.
Figure 9 - Mortality profile of the Mammuthus columbi from Hot Springs, South Dakota (MNIc according to cheek teeth : 36).
35Mammoths, like elephants, must have been gregarious animals. Among current elephants, the herd, led by a matriarch, can include up to thirty individuals from the same family, mostly females with their young, young and old males. Males having reached sexual maturity for two or three years are excluded for their herd and adopt a solitary life, but eventually gather between them, returning to the females only for procreation (Laws 1966).
36In North America, most mammoths assemblages have a type B profile resulting from selective mortality affecting a mixed herd (this is the case for sites with Clovis points, fig. 1) or a type C one resulting from a selective mortality affecting only males or a non- selective mortality affecting a declining population (Haynes 1991). Type C is for example found on the paleontological site of Lamb Spring in Colorado where forty Columbian mammoths were unearthed (Stanford et al. 1981).
37Mass mortality affecting a mixed, stable and growing (type A) population has the specificity to mainly include the very young or juveniles (nearly 50% of under 12 years old) as well as females. As for the other age classes of type A, their number decreases progressively with increasing age (Haynes 1987). This profile usually indicates a catastrophic death resulting from an epidemic, a flood, excessive drought or fire.
38Type B profile is characterized by an extremely large number of very young and juveniles (over two-thirds are under 12 years). A few young adults of approximately 13 to 24 years and some individuals at the end of their life may also be represented. However, adults and older adults, about 22 to 48 years old, most likely to thwart the attacks of predators or the traps of Men, are hardly present.
39As for the type C profile, it would result of a non- selective mortality on a declining population or a selective mortality that affected only males. The main victims are adults between about 25 to 38 years old, followed by young adults approximately 13 to 24 years old and adults aged 37 to 48 years old. Very young individuals and juveniles under 12 years of age, as the elder over 49 years old, are therefore underrepresented. Type C profile, which may correspond to a catastrophic mortality occurring in a declining population, shows a minimum of subadult individuals.
40The mortality profile obtained at Hot Springs does not correspond precisely to any of the main types observed by G. Haynes (1987).
41The first interpretation, put forward by Dr. L.D. Agenbroad and his team, is that the trap acted on solitary males or male couples but would not have concerned any family made of females accompanied by their young. Indeed, the lack of deciduous teeth (D2, D3 or D4) confirming the absence of young individuals might suggest that the trap did not reach family groups.
42This mortality profile has very few equivalents, both at a paleontological point of view and in an ethological perspective. However, with the help of Gary Haynes it was possible to draw a parallel between the profile obtained in Hot Springs and the assessment of the proportion of males he drew from populations of Loxodonta males from different National Parks in north-western Zimbabwe (personal communications 2005 and 2008, Haynes 1991). Although his estimate of ages is done approximately, it is clear from numerous observations that there is a neat overrepresentation, in males, of elephants in the prime of age. The 13-36 year-old total well over half of the individuals, followed to a lesser extent by the under 37 years old accompanied by an absence of individuals under 12 years of age. These estimates in proportions of male elephants put forward by Gary Haynes only take into account here the males that were not integrated into mixed herds. They may underestimate some individuals of more than 37 years if we consider that adult males of approximately 37 to 48 years can be temporarily associated with the mixed herds even of they remain primarily composed of females and young (Haynes 1991).
43The overrepresentation of adult mammoths between 25-38 years, nearly two out of three in the case of Hot Springs, the low proportion of other age groups and the absence of individuals under 12 years old are as many common points between Hot Springs and the sites mentioned above. The interpretation of this mortality profile is difficult to establish, especially as such a configuration is not sufficiently documented at the moment, as we just mentioned. However, the strong similarities observed between the profile of Hot Springs and the proportions of males observed in Zimbabwe would tend to compare this profile with the type C defined by Gary Haynes, resulting notably from selective mortality that affected only males, except that in both cases subadults are absent
44Except the site of Hot Springs, which is an exception, most deposits with high bone accumulation of Elephantidae (fossil or modern) usually result mainly from mixed herds containing at the most several adult males (Haynes 1991).
45If we look at the Asian elephants, Elephantidae are excellent swimmers and the adults do not hesitate in putting themselves in danger to help their descendants. It could be imagined that the adult females at Hot Springs could have rescued their young, carefree and vulnerable, by helping them to get out of the water. The fact that there is no young does not necessarily mean that the trap affected only isolated males; it may also include members from groups of females out of which the youngest, less heavy, could have escaped.
46This hypothesis could possibly be supported by the overrepresentation, in number of remains as well as in MNI, of pachyderms about five to ten tons at the expense of smaller mammals (giant short-faced bear of nearly 600 kg, camels, llamas, pronghorn antelopes, wolves, coyotes, badgers - Agenbroad, Mead 1994) and of young mammoths still far from their maximum weight. Indeed, if we judge by present-day elephants, the weight of the young mammoth at birth should have been close to 100 kg, and then gradually increased at a rate of a kilo per day, reaching an adult weight of around eight tons. Thus, it appears that Elephantidae trapped in Hot Springs would not weigh less than 4.5 tons.
47The isotopic oxygen study carried out by D.C. Fisher on samples of tusks from Hot Springs seems to indicate an annual “bimodal” distribution of mortality. The trap could have worked especially in spring and autumn (Fisher et al. 2003). The distribution of mortality could thus be explained by the particularly unstable soaked banks in both seasons, while the dry soil in summer and the frozen one in winter would have facilitated the ascent of elephants by more stable banks that did not give way under their colossal weight.
48Finally, even if too few bone elements have allowed to establish the sex ratio of the population, females are not excluded from this accumulation as evidenced by one of the most beautiful specimens in connection in situ (postcranial skeleton 13S23W) assigned by A. Lister to a female (Lister in Agenbroad, Mead 1994). This calls into question the previously mentioned hypothesis that evokes a possible lack of family groups in the composition of the bone assemblage of proboscidea.
49The contribution of the analysis of the Hot Springs Elephantidae cannot be overlooked because it completes unexpectedly the existing natural mortality referential. However, it is difficult to interpret accurately the resulting mortality profile of Mammuthus columbi. Indeed, a significant over-representation of adults and the absence of young animals are evidenced out of a minimum of thirty- six individuals, taking into account the analysis of the molars. This type of configuration has only few equivalents and remains insufficiently documented at present. Therefore, no interpretation is preferred. The trap could have affected occasionally isolated males, small groups of males, or mixed herds out of which the younger, less heavy, could have extracted themselves with the help of the adults. Conversely, it is possible that unstable banks collapsed under the weight of huge adult pachyderms, leaving them little chance to get out of this mess which then offered no more hold in case of fall, otherwise said no escape.The summer excavations led by L.D. Agenbroad and his team could help to update the MNI with the latest findings and to clarify the mortality profile within a few years.